Building a National Knowledge Infrastructure: How Dutch Pragmatism Nurtures a 21St Century Economy

tHe C oo k re p ort on internet p rotocolBuilding a

BuildingBuilding AA nnaationtionaall KnowledgeKnowledge IInfrnfraastructurestructure

HowHow DDutcutchh PrPraagmgmaatismtism nurturesnurtures aa 2121stst CCenturyentury EEconomyconomy n a tion a l Knowledge I nfr a structure

The COOK Report ISBN 978-90-814102-2-9 On Internet Protocol What is the COOK Report?

The COOK Report appears 6 times a year with exclusive interviews and commentary on strategic planning for the Internet policy, economics and infrastructure. Subscribers enjoy special member privileges in the private Economics and Architecture of IP Networks mail list. A unique community of interest offering a unique online discussion where industry leaders share knowledge and develop ideas.

Membership is limited to 200 industry visionaries from commercial players such as Google, Cisco, BT, Motorola, Dell, Sprint and others. Other members represent the research and university community, supercomputing centers, network operators, industry analysts and VC communities. Also represented are community network and wireless players, both regulatory and regulatory attorneys, entrepreneurs, public sector, university researchers. 20% of member spaces are reserved for industry executives who work outside the U.S. There is a strong focus on the economic impact of open access fiber, advanced networks and other forms of rich connectivity.

Over the past 18 years special guest interviews have included Vint Cerf, JP Rangaswami, Nico Baken, Hernam Wagter, Carlotta Perez, Chris Locke, (the clue train manifesto interview) John Curran, Simon C Lin, Jerry Michalski, Harvey Newman, Bill St Arnaud and many others. The COOK Report has been using the mail list/forum approach to community building for 8 years. Publisher Gordon Cook was formerly served in the Congressional Office of Technology Assessment, and prior to that, the John von Neumann National Supercomputer Center at Princeton. He also has an earned Doctorate in Russian history from Duke University, 1972.

For more information visit http://www.cookreport.com/

Cover photo: Roos Aldershoff Fotografie MAASTRICHT, THE NETHERLANDS: Inside the Selexyz bookshop, Maastricht. Merkx+Girod architects were asked by the Dutch booksellers Selexyz to convert the interior of the former Dominican Church in Maastricht into a modern bookstore. The incredible finished result not only offers a huge range of books but also gives a very different shopping experience!

This book is produced and distributed by and under license with Cook Network Consultants Building a National Knowledge Infrastructure

How Dutch Pragmatism Nurtures a 21st Century Economy

Gordon Cook

The COOK Report on Internet Protocol

SURF Utrecht The Netherlands 2010 Dedication

To Valerie, Gordon III, Annalee, and Max my family whose support made this possible

and to Carlota Perez whose magnificent work provided the missing link and thereby enabled the connection that provides the foundation for this book

with warm appreciation

Gordon Cook

ii THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY MARCH 2010

PrefacePreface The Origin of Knowledge Infrastructure

The internet and broadband debate broadband – without this I can say As a matter of process I fact– in the US is too close to the from first hand experience I could checked what I was finding with “ground” to enable most observers never have accepted the assign- the folk I interviewed in Portland. to understand where the world will ment from the Netherlands that They answered my questions, be in five years. That reality five turned a double issue of my news- liked my drafts and wanted them years off is the knowledge infra- letter into a book being able to to be read by more than just my structure being built in the Nether- communicate with Europe instan- subscribers and hence asked me lands and described in this book. taneously 3000 miles and six time to collaborate in turning them into This far-sighted work will be trans- zones to the East and with my Edi- the resulting book. formative there in the same way tor Sheldon Renan in Oregon 2000 that water power and steam power miles and three more times zones This book’s intent is to lay out a transformed England in the 18th to the West in the space of 10 roadmap that explains to readers century. This while the rest of the days. how the Netherlands came to its world is arguing over the private national consensus to invest in share-owner model for what I have been reporting on SURFnet national knowledge infrastructure. should be public infrastructure. since Bill St. Arnaud introduced And show what the evolution of the This debate deprives it of the most me to Kees Neggers at a CANARIE resulting policy process has ac- important advantage of optical Workshop in Montreal in 1999. complished. It will conclude by technology. When it makes this My question to Bill was who is do- pointing out the changes in per- transition it will be able to begin ing this right? Who has a philoso- ception that must be accomplished where the Dutch started in 1997 phy similar to yours? His re- in the united states in order to with the parliamentary motion fa- sponse – “Kees Neggers” and the render feasible what Holland has voring the building of knowledge in-person introduction was ac- done. Namely treat the fiber and infrastructure. Meanwhile the complished. I have been report- ducts as the national highway sys- Netherlands has a 13 year lead. ing on SURFnet ever since. At the tem. Those are held in custody As long as the Dutch keep gong end of 2004 and beginning of for the use of the nation and the rest of the world will never 2005 I did a long report. In No- managed by the government which catch up. vember 2008 I met Kees again at services the citizens for the benefit Supercomputing 08 and did a of the people and on which private Nations that don’t understand this much longer report with him and enterprise can build competing will condemn their citizens to be Cees de Laat because I was finally higher layer, profit making opera- the poorly paid peasants of the gaining an appreciation of the tions. underdeveloped world of the 21st very large importance of what century. If the United States they are doing. A year later at Only then can any nation build a doesn’t learn these lessons it will Supercomputing 09 we met again, competitive, economically valuable be transformed into the world’s this time also with Bob knowledge infrastructure. The can leading digital back water. Rather Hertzberger, Hans Dijkman and become possible in the United like the contrast of 19th century Wim Liebrand. States only if competing interests Russia. The intellect and brilliance simply understand the policy prin- and economy was all in either Mos- I found the information necessary ciple of infrastructure as laid out cow or St Petersburg. Elsewhere to answer the questions that came in the chapters of this book. was the largest landmass of back- to mind and wove them into the ward poverty on the planet. The fabric that has become this book. Gordon S. Cook, Jr. US is still blessed. I am fortunate A fabric that was initially intended Ewing, NJ USA February, 3 2010 enough to have some of the better only for issues of my newsletter. iii

Acknowledgments Acknowledgments

To Valerie my wife of 43 years for her good na- following tale. Just as he has been weaving in his tured toleration and love for whom I have become own mind for two decades the concept of ‘netness’ and for being a woman of grace and compassion. or Renan’s law. - Everything wants to be connected and works better when connected. This book is a To my son Gordon III whom I have watched be- specific example of the deeper drivers of Renan’s come a hugely talented builder and shaper of envi- Law. ronments who has been there in ways beyond com- prehension. To Sara Wedeman one of the most creative, determined, talented women I have ever met who To my daughter Annalee who has become a warm lives in the trans-disciplinary mind set that the independent strong loving mother for my grandson world may get to five years from now. Psychologist, “Max”. And for Max who must inherit the digital in- econometrician, down loader of the complete data- frastructure described here in order to thrive. bases of the US Census, and practitioner extraordi- naire of behavioral economics. Visionary proponent To the Colonel, Dave Hughes, West Point Class of of broadband mapping and Amartya Sen who, if ten 1950 also known as the “cursor cowboy’ who has years from now there still is an NTIA, Sara will lead been a good friend since 1980 and who when I fin- the agency to act in the interest of our nation. And ished my policy work at the US Congress Office of an artist who generously devoted more than four Technology Assessment on March 1, 1992 when I full days of design work to get the cover of this book asked what the hell will I do now? said without done right. missing a beat - write and publish a newsletter about this internet thing. Dave saw the synergy and possible with convergence of my skills and the technology. Standing on the cusp of the 19th year of the To Eva Waskell who in the Spring of 1992 discov- COOK Report I have never before been so excited ered that there was someone at OTA trying to de- about the possibilities. scribe this Al Gore research network and invited me to talk at a Tele-strategies conference in April 1992. To Sheldon Renan whom I have known now for And whom, as I only found many years later, 25 almost four years. Man of renaissance talents, years ago began what has morphed into the nation writer, producer, designer, poet, master story teller wide election integrity movement. No matter what and seer. Finder of the rare essence in our sea of the obstacles Eva has never budged. chaos. The spirit who has guided me to connect a carpet of events and ideas that want to be related All visionaries, all determined, all motivated by the want to be connected and catalysed them into the public good. All Bodhisattvas.

Preface iii

Acknowledgments iv

Cook report On Internet Protocol February/March 2010 Section 1 Contents February/March 2010: Section 1, p. 102

Cook report on Internet Protocol February 2009 Section 2 Contents February 2009: Section 2, p. 65

Section 1

Cook report On Internet Protocol February/March 2010

THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010

I. How the Dutch Did It Right When so many others get it wrong

In 1965, French film director Today it is the Netherlands Volume XVIII, No. 11-12 Jean-Luc Godard made a film which has consistently taken February - March noir called Alphaville about a the initiative, made the in- ISSN 1071 - 6327 grim future in which all the vestments and delivered the world’s infrastructure is con- goods. The Dutch have de- much of the Internet — the trolled by a computer. Imag- veloped blazingly fast hybrid Dutch are committed to col- ine that. Towards the end of optical networks of hitherto laboration both inside and Alphaville, the hero pulls the unimagined efficiency and outside the country. They computer’s plug. Suddenly cost effectiveness. have been a proactive force nothing works. The world for international collaboration goes blind. People are feeling They have taken the tech- with scientists and network their way along walls. nique of user controlled light specialists in the EU, the paths as developed by CA- United States and elsewhere In 1965 infrastructure con- NARIE in Canada and incor- with their Global Lambda In- trolled by computer was sci- porated this on a robust plat- tegrated Facility or GLIF. ence fiction. And today? form of Web services and with collaborative e-science In 1965 Today lives and economies middleware into optical net- depend on the Information work technology. So their infrastructure and Communications Tech- networks cannot only trans- controlled by nology (ICT) Infrastructure mit huge data files across more every day. The quality oceans starting one side of computer was of the national ICT infrastruc- the world to the other instan- ture increasingly defines the taneously. But increasingly science fiction. quality of life as well as op- the Dutch can support the portunity for the future. collaborative and multi- And the more you learn, the disciplinary practices that are more it is apparent that the It used to be the United being described as “the 4th Netherlands is building the States, with it’s Internet ini- paradigm”, data intensive electronic network and tiative and “intelligent net- science. knowledge infrastructure on works” that claimed the lead- which the economy of the ership position. Now that’s Unlike the United States — 21st century will be based. the science fiction. where private interests have This report will explain their walled off and Balkanized continuing technology direc-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 1 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 tion -- a direction that on its This is the change described in a never ending pursuit of own level is quite impressive. by Carlota Perez that is technical excellence and ex- common to all technology treme performance. And it is So why is this happening in a revolutions. Speculative or the counter-intuitive strategy country of only 17 million finance capital in the support in which geek values of open people, small and crowded and development of ICT must source software and collabo- with one of the highest popu- no longer predominate. So- rative networks are har- lation densities in the world? ciety must shift to use of vested into creating public/ Why is this not happening in productive capital. In other private partnerships that the United States or Canada words it must use money for evolve into the business op- or China or Brazil? Those are infrastructure to install these portunities and a dynamic all big countries, rich in re- ICT resources in society and and competitive national sources, that are supposed to treat them as knowledge in- economy. “own the future.” This is an frastructure. They become important question. the logical follow on to roads …embrace and highways, canals, rail- We can begin with what any- roads, electric grids, airports disruption and use body who has worked in water and sewage systems technology knows. The real and electric plants. In the the improved problems with technology are end they are simply an inte- productivity not, typically, the technology. gral part of the basic infra- The real problems have to do structure of an advanced and for the public with intentions, models, gov- civilized capitalist nation. good. ernance, implementation, re- sourcing, and user support. John Hagel and John Seely It’s not the technology, it’s …the Netherlands is Brown recently updated their what you do with it. And that, building the sobering 2009 Shift Index: in turn, depends upon what Measuring the forces of long- you have intended to do with electronic network term change, which reveals the technology. knowledge that in the U.S., despite an economic focus on private So when we ask the ques- infrastructure on good rather than public good, tion, why are the Dutch doing American businesses, includ- it right, when so many other which future ing telcos, now earn 75% countries got it wrong, the economies will be less return on assets than answer begins with the fact they did in 1965. One of the that they have better inten- based. reasons, according to John tions and a more inclusive Hagel, speaking at the 2009 process. In demonstrating So, once again, it is not the SuperNova Conference in San admirably farsighted planning technology so much as it is Francisco, is that businesses and negotiated discussion the thoughtful and careful have not been able to ration- among their stakeholders, way that technology policy is alize the disruptive advances the Dutch are leading the determined. It is the way in technology which (thank world in making the ICT that policy is turned into real- you, Moore’s Law) never stop technology transition. ity. And it is the way in which advancing. they continue to push edges

The severity of events is In the Netherlands, the ICT In trying to understand the made more difficult because infrastructure ecosystem emergence of the Nether- of the doctrine that the inde- does have a strategy. It lands as a leader in ICT infra- pendence of the private car- works for them. And it works structure, again, this is not rier is sacrosanct. The prob- for us. And that is to embrace just about network technol- lem is that the direction of those disruptive advances ogy or the network applica- technology has been moving and use the improved pro- tions. Rather the key lesson ever more rapidly into the ductivity for the public good. to take away is that the creation of network capability Dutch network and research of almost limitless abun- The COOK Report contends effort has been built by dance. In contrast to this that the Netherlands, through means of an exceptional at- technology push, the invest- the good fortune of rather tention paid to the economic ment pull of the rules by unique circumstances over impact of the network as in- which the share owner Cor- the past decade, has been frastructure that contributes poration is governed de- able to articulate a vision to the Dutch national inter- mands that management whereby it begins to treat its est. take the opposite course and information and communica- establish a regime based on tion technology investment And beyond that is another scarcity - measured usage, as an investment in public parallel story. It’s the story of constricted bandwidth, con- infrastructure rather than how history and economic stricted user freedom and private share owner deter- circumstance shaped the charging the user the very mined enterprise. The nation character and confidence of maximum that traffic will telecommunications infra- the Dutch. It tells what can bear. The privatized carrier structure is treated as a pub- happen when a nation, becomes a predator that lic rather than a private good. toughened by centuries of feeds on society that in the- Partly this is for the purposes challenges that would (liter- ory it serves. The result is an of science and research. But ally) sink most countries, enormous gap between the it has allowed the country to learn how to work together to technological capabilities at- develop world leading tech- leverage technology at the tainable with state-of-the-art nology that operates along- infrastructure level and over- optical network technology side the share owner main- come impossible odds. It’s a and the reality enforced by tained voice network of KPN great story, and it starts in share owner networks in their as well as those of the MSOs Chapter II. respective societies (Cable TV companies) of the Netherlands.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 3 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 A Question for the United States: insurers suffer in common under what might be termed ‘the competitive death embrace’ ". How to Make Beneﬁts of Thinking Long-term More Visible? The death embrace is best illustrated by the thinking "If I don't do this marginal deal, my competitor The long-term vision and achievements of SURFnet will, so I should do the deal rather than let the and the Dutch infrastructure ecosystem call into benefit go to them." If you review the papers, you'll question the failure of U.S. infrastructure players to see countless variants of that statement, and if you be equally committed to the long view. One ques- pay attention, you'll notice that not one says "bene- tion we need to ask, as a nation, is this. How do we fits and risk". make long-term risk and benefits more visible so that people and institutions can make better This problem is compounded by the short-term bi- choices? ases of corporate securities reporting. A Warren Buffet understands this very well and reads through Failure to understand true long term risks and op- it. (Many readers of the Cook Report understand portunities is responsible for many investment cycle this but most lack the investment discipline to profit failures. Most recently the mortgage crisis, but more from it.) commonly the cycle in which insurance companies lower their rates in lock step to stay in business, In technical contexts, we see companies and deci- only to find during some catastrophe that their un- sion -makers repeatedly caught in what Harvard’s derwriting models were too optimistic. On investiga- Clayton Christensen has termed the "Innovator's tion, they invariably find that (a) the models really Dilemma." weren't very good, but (b) the risk assessment as- sumptions in the models had been progressively http://en.wikipedia.org/wiki/Disruptive_technology downgraded in response to competitive pressure. This is not a consequence of stupidity. It is a re- Disruption, whether in markets or technologies, is quirement for market survival. Or to build a great inevitable. Too many choose to ignore that fact. The infrastructure. cause can ultimately be traced to the fact that quar- terly behavior can be explained to investors while Johnathon S Schapiro wrote to the IP list on Jan 4: long-term behavior can't. “While bank lending practices were, in my opinion, largely responsible for the mortgage crisis, it must be acknowledged that banks, trading houses, and

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 4 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 II. The Netherlands National ICT Research Infrastructure History, character, policy and pragmatism

Infrastructure and the for stable industrial relations, National Interest … after the war fairly low unemployment and inflation, a sizable current infrastructure account surplus, and an im- After World War Two, gov- portant role as a European ernment in both the United had to be rebuilt transportation hub. Industrial States and Europe continued activity is predominantly in the social contract that had from the ground food processing, chemicals, helped end the problems of up. petroleum refining, and elec- the Depression. In Europe, trical machinery.” however, it was necessary to http://en.wikipedia.org/wiki/ rebuild infrastructure from The Economy of the Economy_of_the_Netherlands the ground up. The flourish- Netherlands ing of free market capitalism was kept in balance by a Wikipedia says “On the Index Role of Natural Gas in generally accepted consensus of Economic Freedom, the Dutch Postwar Good that there needed to be a so- Netherlands is the 13th most Fortune cial safety net, and that gov- laissez-faire capitalist econ- ernment needed to create the omy out of 157 surveyed The Netherlands has, how- conditions under which its countries. At the time of writ- ever, an asset that for a lim- citizens had basic food, shel- ing the Netherlands is the ited time continues to gener- ter and education. 16th largest economy of the ate money for investing in world. Between 1998 and social capital. The Nether- The Dutch have done an ex- 2000 annual economic lands is presently the world's traordinary job of determin- growth (GDP) averaged fifth-greatest natural gas ex- ing how to leverage their re- nearly 4%, well above the porter. According to Wikipe- sources in order to create a European average. Growth dia, “While its oil reserves in knowledge infrastructure that slowed considerably in 2001- the North Sea are of little im- will serve the national inter- 05 as part of the global eco- portance, the Netherlands is est, support EU partnerships, nomic slowdown. 2006 how- presently the second-greatest and compete successfully in ever, showed a promising [after Norway] natural gas the international arena. 2.9% growth. Yearly growth producer in the European Un- Meanwhile in the United accelerated to 4.2% in the ion and the ninth-greatest in States, infrastructure discus- third quarter of 2007. Infla- the world, [after Russia] ac- sions and investments have tion is 1.3% and is expected counting for more than 30% been sporadic. The U.S. has to stay low at about 1.5% in of EU total annual gas pro- not pursued a serious infra- the coming years. The Neth- duction and about 2.7% of structural agenda since the erlands has a prosperous and the annual world total. interstate highway system open economy, which de- Proven natural gas reserves was constructed in the 1950s pends heavily on foreign of the Netherlands are esti- and 60s. trade. The economy is noted mated (as of January 2005)

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 5 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 at about 50-60 trillion cubic national infrastructure of the then defeated the mighty feet, or about 0.9% of the Netherlands, while 85 percent Spanish armada. world total. Although the went to welfare benefits, in- http://en.wikipedia.org/wiki/ Netherlands owns substantial terest payments on the na- Cornelis_Corneliszoon gas reserves in the North tional debt, and spending on The COOK Report contends it Sea, most of its production is health care, education and is this heritage that has en- presently from on-shore public administration.” abled the Dutch to begin wells, and much of the natu- their critical and innovative use of the gas funds in build- ral gas produced by the “Ruud Lubbers, who was ing the knowledge infrastruc- Netherlands comes from economic affairs minister in ture that the remainder of Groningen Province, which the seventies and prime min- this chapter will describe. borders the North Sea. ister from 1982-1994, com- [Same Wikipedia article cited plained that the money led to above.] a "lack of discipline". It On the Other Hand the wasn't until the late eighties United States Plies a Drilling began in the that Lubbers saw an oppor- Different Course Slochteren natural gas field - tunity to create a fund for one of the biggest in the structural improvements to In the United States, the world on July 22, 1959. Fifty the economy. The "gas fund" Rooseveltian New Deal put in years later according to a (Fonds Economische Struc- a framework of financial June 17 article in NRC Han- tuurversterking, FES) finally regulation that enabled the delsblad “the total revenue saw the light in 1993.” U.S. to prosper after World from the Groningen gas field War Two. In the 50s, under is more than 211 billion In many countries, revenues from natural resources in- President Dwight David Eis- euros, most of which went enhower, national prosperity straight to the treasury.” variably are siphoned off by increased during a period http://www.nrc.nl/internation private interests. The Dutch have had the discipline to when the top income tax rate al/article2274261.ece/The_D was what today would be an utch_curse_how_billions_fro turn the income from gas to infrastructure and then to unthinkable rate of 93%. m_natural_gas_went_up_in_ Under President Kennedy, smoke knowledge infrastructure. The however, that rate was cut to “The Slochteren natural gas difference may be the innate pragmatism of the Dutch roughly 70%, The course of field also turned out to be the next four decades was much bigger than expected. character. For centuries they have faced unending chal- increasingly set by the in- Back in 1963, the NAM esti- creasing triumph of the Chi- mated reserves at 1,100 bil- lenges that have required cago School of free market lion cubic meters. That num- working together and managing a commons involving wa- economics which preached ber has been revised up- that economic growth should wards to 2,700 billion cubic ter resources and innovative technology. Everyone knows be achieved by more and metres, more than 60 per- more tax cuts. This resulted cent of which has been ex- the story of the dikes and in increasing disparity in fi- tracted. . . . “more than 52 windmills. Fewer know of the breakthrough invention of a nancial reward for the work billion Euros, or almost a force. CEOs salaries grew quarter of all natural gas windmill-powered sawmill that enabled the Dutch to from a multiple of 20 times revenues, went into financing entry level worker earnings social security. Only 15 per- build a flotilla of smaller, to a multiple of several hun- cent was used to improve the faster, and cheaper ships that

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 6 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 dred times entry level com- manufacturing hugely de- Perez points out, economic pensation. In 1968 the CEO clined. The manipulation of policy decisions made in the of AT&T made $200,000. money made up an unprece- national interest, as opposed Today chief executives at dented share of U.S. GDP. to supporting Wall Street, AT&T and Verizon receive can result in disseminating salaries of up to $30 million. Ironically this phenomena is technology and its benefits After bonuses, their annual sometimes described as “the across the broadest reaches incomes may exceed $100 Dutch disease.” The term was of society. million dollars per year. coined by The Economist in 1977 to describe how the dis- See: …challenges covery of natural gas in the http://en.wikipedia.org/wiki/ Netherlands resulted in a sig- Carlota_Perez required the Dutch nificant decline in manufac- and to work together to turing there. That said, the http://www.carlotaperez.org/ Netherlands learned how to Articulos/TRFC-TOCeng.htm manage a commons manage it. The United States and did not. http://www.cookreport.com/c …with innovative http://en.wikipedia.org/wiki/ arlota.pdf technology. Dutch_disease But this is not what happened in the U.S. and Britain. Rea- The dominant ideology in the …in the U.S. gan was elected in the U.S. US became that the govern- private interest at the same time as Thatch- ment was “bad” and needed erism triumphed in the to be starved. On the other was put before United Kingdom. The incum- hand, success of private in- bent carriers were privatized dustry and the so-called free public good. along with the early cable tv market “would raise all systems where growth in the boats”. In reality it meant As the integrated circuit (in- EU was far behind the US. that the American govern- vented in 1958) matured and With support of incumbent ment had limited ability to made possible the global telcos, economic market con- act in the national or public internet, speculative capital ditions were established to interest. When given a choice funded that growth in the obstruct effective use of new between funding infrastruc- United States, giving it sig- optical fiber technologies. ture or funding the military, nificant control of the na- military always won. This tional economic agenda. The The concept of a general pur- trend reached critical mass result was the extension of pose technology (GPT) is during Ronald Reagan’s what Carlota Perez in her useful in understanding this presidency. It enshrined the seminal 2003 study of tech- issue. GPTs can affect an supremacy of private interest nology revolutions defines as entire economy (usually at a over public good. This re- “the transition phase” needed national or global level) and sulted in significant increases to shift national policy from have the potential to drasti- in the national debt as well as necessary investment in pro- cally alter societies through the financialization of the ducing a new technology to their ability to disrupt pre- economy. Growth was fueled the more efficient use by na- existing economic and social primarily by credit, while tional government of produc- structures. productive activities like tive capital. In this state,

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 7 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 …incumbents tive mail hauling franchises, of early packet data networks etc. Electricity and tele- and the need for having a obstruct effective phone were done primarily as coordinated government ap- use of optical fibre regulated franchises (until proach to the creation of a the new deal when it was re- university and research- technology alized that a large part of the oriented infrastructure. As a nation would never be served result the SURF foundation so we set up cooperative and was created along side the Mark Cooper, Director of municipal entities). Free NWO or Dutch Science Foun- Research for the Consumer of Federation of America wrote market ideology still domi- dation. nates U.S. policymaking. Un- on January 8 2010 about the til we have overthrown that “The SURF Foundation was potential benefits of dissemi- ideology, there is no way to established in 1987 to co- nation GPTs in the United fund a ubiquitous, adequate ordinate the implementation States. “The extremely dif- fuse benefits mean that the national broadband infra- of a multi-year plan for the structure.” improvement of the applica- dominant funding mechanism tion of information technol- of the past thirty years (pri- ogy (IT) in Dutch universi- vate capital) won't work very The SURF Foundation ties, schools for higher voca- well. The approach to funding the last three GPTs in the As privatization of KPN, the tional education and research institutes. In the course of its U.S. was very mixed and Dutch incumbent, ap- activities SURF has become a pragmatic. Railroads were proached, and small urban nationwide supplier of serv- first funded in the private cable systems were sold off, ices. These services are pri- sector and they all went far-sighted people in the bankrupt. We then funded Netherlands were quick to marily provided through its operating subsidiaries: them with land grants, lucra- appreciate the development SURFnet bv and SURFdien- sten bv. SURFnet manages the computer network of the same name. SURFdiensten deals with licensing agreements in the fields of software, hardware and information services.” http://www.ariadne.ac.uk/iss ue5/surf/

The SURF site itself says: “SURF is the collaborative organization for higher education institutions and research institutes aimed at breakthrough innovations in "On 13 June 1986, RARE (Réseaux Associés pour la Recherche Européenne) is ICT. SURF provides the foun- established as an association under Dutch law by Hans Rosenberg on behalf of the University of Utrecht and Klaus Ullmann on behalf of the DFN Asso- dation for the excellence of ciation. This photo and caption is found in"Terena Celebrating 20 years" higher education and re- booklet: http://www.terena.org/publications/ﬁles/20th_anniversary.pdf search in the Netherlands.”

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 8 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 http://www.surffoundation.nl the research and higher edu- as a company in cooperation /en/oversurf/Pages/Default.a cation community. with the PTT (at that time the spx state owned telecom monop- …Hans Rosenberg oly operator and since 1989 Kees Neggers, SURFnet Di- privatised into KPN), the rector says that “SURF ex- understood why a government also funded the isted already [before 1987] small country start of SURFnet as a perma- as an informal approach to nent organization.” cooperation between the Uni- needed a big versities with regular meet- Leading the initiative was the ings in which the University network. visionary Hans Rosenberg, he Board member responsible was member of the board of for ICT and the Director of The Government reacted the University of Utrecht. Be- the Computing Centre par- positively to this informal ini- fore that he was an alderman ticipated.” It was this group tiative by first financing the in Utrecht and was prominent that took the initiative in the start of SURFnet as a project in solar radio astronomy. Ro- mid eighties to start a na- and two years later, under senberg was the founding tional research network as the condition that SURF father of the current ICT in- part of a broader ICT plan for would first become a formal frastructure in the Nether- legal entity to create SURFnet lands. He helped start SURF,

The Development of AMS-IX other ISP's were also allowed to connect and the name AMS-IX was first used. In 1997 NIKHEF and SURFnet handed over the formal responsibility of The Science Park Amsterdam (the relatively new the distributed exchange to the AMS-IX Association. name of the location of the networking institutes In 1990 the Association formed the AMS-IX limited and AMS-IX and Netherlight) in is a historical net- company, AMS-IX B.V. The Association held all working hub. By 1982 CWI the National Research shares while all assets are transferred to the com- institute for Mathmatics and Computer Science had pany AMS-IX B.V. Until 2002 SURFnet continued to become a European hub in EUnet with its famous manage the overall operations of the exchange. “mcvax.” In the early nineties, NIKHEF, with Rob Blokzijl as its driving force, and SURFnet at SARA Today AMS-IX still operates as a not-for-profit formed a layer-2 shared infrastructure to exchange community based exchange with well over 300 traffic between (academic) organizations. In Feb- connected members. With a daily average traffic ruary 1994, it was internationalized as a commu- over 500 Gbit/s and peak traffic over 800 Gbit/s nity based Internet Exchange Point to exchange AMS-IX today is the largest internet exchange in traffic with CERN in Switzerland. At this point the world.

NetherLight is a SURFnet owned and operated advanced open optical switching infrastructure and proving ground for network services optimized for high-performance applications. NetherLight has been operational since January 2002 with the installation of a 2.5 Gbit/s trans-Atlantic lambda between NetherLight and Star- Light in Chicago. NetherLight is a SONET/SDH cross connect and Gigabit Ethernet switching facility for high- performance access to participating networks. Ultimately, it will become a pure wavelength switching facility for wavelength circuits as optical technologies and their control planes mature. NetherLight is made possible by SURFnet in the context of the GigaPort project. NetherLight is located at SARA in the Science Park Amster- dam, and has been realized by SURFnet in the context of the GigaPort projects. Today over 300 Gbit/s of international lightpath capacity is connected to NetherLight. http://www.netherlight.net/ SARA SARA Computing and Networking Services is an advanced ICT service center that supplies – since more than 30 years – a complete package of high performance computing & visualization, high performance networking and infrastructure services. Among SARA’s customers are the business community and scientific, educational, and government institutions. http://www.sara.nl/aboutsara/aboutsara_pr_eng.html

SARA’s product portfolio consists of: • High Performance Computing & Visualization: facilities and services in supercomputing, data storage, visualization and virtual reality; • High Performance Networking: design, installation and management of advanced Wide Area Networks; • ICT Services: housing and management for third parties of critical infrastructures, systems, applications

The above diagram shows the January 2010 international connectivity at NetherLight, the GLIF Open Lightpath Ex- change (GOLE) in Amsterdam. All other circles in the diagram are also GOLEs like NetherLight, the clouds are the networks connected to NetherLight that are capable of supplying lightpaths. Details of all GOLEs can be found on the GLIF website under Resources: http://www.glif.is/resources/

SURFnet, and was also Foun- completed, and it was con- On January 5, 2010, the 200 der of RARE. And it was sidered not correct to have members of the COOK Report Hans Rosenberg who under- KPN reacting to request for mail list, including 20 mem- stood why a small country tenders from a company in bers who live in the Nether- needed a big network. Ro- which it owned 49% of the lands were queried as to why senberg died untimely in shares). SURFnet had rarely been a 1992, but his ideas and influ- subject for discussion. Was it ence are still felt by the cur- On 31 December 2001 KPN irrelevant to everyday con- rent leadership of SURF sold its shares in SURFnet to cerns such as the fiberization SURF and the next day of Amsterdam? Or was it be- SURFnet started as a project SURFnet and GigaSURF cause it was just unknown? in 1985/1986 Stichting SURF merged and GigaSURF was Or was it some unknown fac- was created in March 1987. abandoned. Since then, tor? In January 1988 SURFnet be- SURFnet bv is a 100% sub- came a not-for-profit com- sidiary of SURF.” The replies were illuminating. pany 'in formation' (which means SURFnet operated as …SURFnet converts Rudolf van der Berg: see: a company, but was not yet a http://internetthought.blogsp formal legal entity, SURF was natural gas into ot.com/ a Telecommunications con- formally responsible for its ICT infrastructure business.) sultant ICT consultant wrote:

In January 1989 SURFnet be- Transition to and “Hendrik Rood or Jaap van came a formal legal entity, Growing up with Till are probably much more versed in the ways of SURF- SURFnet bv, with SURF hav- Infrastructure ing 51% of the shares and net than I am. I personally KPN 49%. (One of the rea- don't see SURFnet as such an SURFnet is important for enormous accomplishment. sons for the delay was that it several reasons. The most Given the size of the Nether- was only in January, 1989, important are its consistent lands and given the nature of that PTT became KPN as a services and persistent visi- the demands I wouldn't know private company. Before that bibilty – and especially its it would have been more how you would deal with in- role in enabling the funding terconnecting academic insti- complicated for the PTT to of the ICT infrastructure via tutions differently. If any- take shares in a company). natural gas revenues in the thing what helped was that 1990s. That SURFnet and the Netherlands realized that In 1999 GigaSURF bv was related entities like AMS-IX academic networking was started, 100% owned by and NetherLight are just important earlier than others SURF, to do the request for “there” now and taken for and by keeping that lead, tenders of the SURFnet5 granted is perhaps the single make the Netherlands the network as part of the first largest accomplishment that default gateway to Europe. GigaPort project, SURFnet the SURF Foundation has Stuff like lightpaths just fol- continued to do the actual made during the last 15 lows from the technology and work. (This construction was years. isn't really that flabbergast- needed because KPN initially http://wiki.glif.is/index.php/N ing. That said if you compare was unwilling to give up its etherLight shares after the liberalization SURFnet with the rest of the of the Telecoms market was world, we somehow seem to

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 11 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 have done right or better said couple of folders publicly volved with the establishment much of the world seems to within the campus and some- of the campus wireless net- have done wrong. But let me one build a cool search interwork, my first employer the explain why I feel it is such a face to make it all accessible. local internet exchange natural thing to do and why I The network became an (NDIX), the Trent regional do believe SURFnet was in- enormous breeding ground telecom network and I dare volved in building these for new ideas and network say even with some of the ideas. use. thinking behind FTTH in the Netherlands. For instance “SURFnet meant Having joined that group in some of Reggefiber's people 1998/99 and becoming vice had an office in the Virtu Se- there was chairman of the campus net- cure Webservices (now work, I only learned about Equinix) building where the always enough networking in the Ethernet/IP NDIX was located as well. We bandwidth...” way of thinking. Traditional were discussing FTTH then RUDOLF VAN DER BERG networking based on circuit- and some of us are still in- switched or smart network volved with it now. SURFnet ideas have always felt erro- wasn't involved there, though I came online in 1994 at neous to me. Whenever we connections to SURFnet were Twente University where my played with new technologies one of the reasons why cus- current group of friends or- (like streaming HD video on tomer owned regional fiber ganized as Datanet Drienerlo demand in 1999/2000) we took off in the East of the (DND) were rolling out a did not need to look at the Netherlands. 10 Mbit/s Ethernet network network and looked at the on campus and 100 Mbit/s service instead. When a com- “…people just switched was available by pany asked us to test drive 1998. At first this was an un- their HD video on demand don’t realize official non-university sanc- server, some friends rebuild tioned network built with their two years of work within infrastructure is whatever was available and a a day and then some tweak- crucial until things good drill to get a cable from ing for a week. There were one dorm to the other. From lots of questions whether go wrong...” 1994 Twente University paid Linux or the machine could LEO PLUGGE and rolled out the network, handle the task, but no one but the students operated the asked if the network could No matter which way you network. The campus net was handle it. It was assumed it look at it. SURFnet wasn't a and still is hooked up to could despite all traffic of the direct driving force behind SURFnet with at first an ATM students accessing movies much of what happened in uplink, but later on GigE and across the network on hard the Netherlands, but it was 10 GigE. The students have disks on other PC's across involved or a known quantity been demanding customers campus which would compete in the background. SURFnet ever since the Campusnet with the streaming VOD for allowed a way of thinking started. Much legal and illegal bandwidth. that assumed there was al- traffic was handled on the ways enough bandwidth network, but for instance DND is still active after more available and it often helped standard P2P failed to catch than 20 years and can now out innovative parties with on as everybody shared a be credited to have been in-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 12 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 that bandwidth. Once that hard way after a fire in their benefit become detached to way of thinking is with SURF- computing center in Novem- the problem and the solution. net as part of your telecom ber 2002. Or take New Or- As long as there is no major genetic make up, it becomes leans as an example. But disaster/disruption/shortage, normal to think in GigE since this was infrastructure, there is no need to interfere uplinks instead of T1's. I do as Kees has pointed out “To- or worry. The same goes for believe that SURFnet allowed gether with BT, Cisco and the electric power provisioning). a whole generation of net- University SURFnet were able work nerds not just at to install a complete new Luckily, we had a few vision- Twente, but at every univer- SURFnet node at another lo- ary people come-up with the sity who in their formative cation on the campus, includ- idea to create SURF in 1985 years in university knew just ing a new GSR router and to realize SURFnet, and we one thing: "The network can fiber connection, which be- now have the government handle it". came operational the next decision to treat it as an in- day after the fire.” It’s not frastructure of national im- Leo Plugge, Secretary of the unlike our water manage- portance. SURF Advisory Committee: ment infrastructure that is an “It’s interesting to see how example to the rest of the Herman Wagter (who is SURF’s infrastructure is taken world, but is taken for building Citynet in Amster- for granted - or even as granted by most Dutchmen. dam): As proud as we may natural - by many users. As Without a tea party to under- be, there are more self- Rudolf wrote: “I wouldn't score the importance of infra- organized institutions in the know how you would deal structure, we run the risk of world who have been long with interconnecting aca- dozing off, as we did before lasting, even longer than the demic institutions differ- the 1953 flooding, which was Waterschappen. ently”, and students know a very hard wake-up call in- "The network can handle it". deed. The last Nobel Prize winner Elinor Ostrom has written …government Is it a national asset? Yes. about them in a commend- Our current IT-infrastructure able (very readable) book, decision to may seem a natural thing to Governing the Commons, treat it as an do, but that does not mean that shows how classic eco- that it’s not unique in its nomic science has ignored infrastructure of kind. Take our water de- the potential for self- fenses. The fact is, that the organization, assumed that national Dutch organizational and the “tragedy of the com- importance… technical infrastructure arose mons” is a basic law (which it naturally from villagers who isn’t). organized themselves locally Infrastructure is essential, into Waterschappen since the “…it’s no accident but it tends to be hidden in 12th century. This was natu- daily life. Only when things ral, but also unique in the SURFnet matured go wrong people (users) be- world. into a major come aware how crucial infrastructure is. As the Univer- Having delegated the respon- infrastructure.” sity of Twente learned the sibility to solve the problem VINT CERF also means that those who

some infrastructural manure, team since its origins and Jaap van Till: AMS-IX, SURF water and sunlight. have worked closely, from and SURFnet are in a sense time to time, with Kees Neg- cases of "Triumph of the Steve Wolff: However natu- gers and others there. Commons". Well understood ral SURFnet may seem to its Frankly, I have been ex- of what resources/services clients and customers, it has tremely impressed by Kees are for shared interests (non undeniably been from its very and his colleagues. Their abil- combat) and definition of beginnings a technology ity to be in the center of very what not shared (competi- leader in the worldwide NREN high capacity networking active). That NetherLight, community. In addition - at tivity and be reliable provid- GLIF, GLORIAD, and in a cer- least from an outsider's per- ers of super high speed plat- tain sense 'the national layer spective - it has been a forms is a significant accom- two SURFnet optical network' model of how to deal with its plishment and certainly put are not well known in the national government in pro- the Netherlands at the lead- Netherlands is that these are viding services and avoiding ing edge of academic net- big-pipes for a special group the fate of those NRENs (Na- working. If their effort has of big-science like nuclear tional Research and Educa- now matured into a national data of CERN and long base- tional Networks) who are in- infrastructure, it is no acci- line radio astronomy. creasingly seen as competi- dent. tors of unsubsidized private These applications need so ISPs. Exchanging a tempo- much bandwidth capacity rary resource (and have limited budgets as [Editor’s Note: Steve Wolff always) so they have been explained to me later the for a sustainable taken off the Internet con- same day in a voice conver- knowledge infrastruc- nections. Otherwise the rest sation that far too many ture of the internet in Europe NRENs are seen by their uni- would have grinded to a halt versity CIOs as means of get- A paper by the Netherlands during these trans-national ting subsidized broadband for Bureau for Economic Policy scientific experiments. It was web and email for students Analysis gives an excellent recognized long ago here that and faculty and have precious overview of Dutch budgeting it is useless policy that by little to do with enabling ap- philosophy and the thinking thinking (vertical) of new plications not possible on that undergirds the use of services you can stimulate commercial networks.] gas money for infrastructure. network use. It does not Readers will see that it ends work like that. Scalable net- One COOK Report member with great uncertainty about work infrastructures ( in hori- who responded was not a the sustainability of the zontal layers) will boost in- citizen of the Netherlands, Dutch economic course. (The vention of new applications but he does know something COOK Report does not share by removing obstacles to about SURFnet. He is the that uncertainty.) growth. man many call “the father of The Internet” and currently From The Dutch Fiscal Good roads create traffic. serves as Chief Internet Framework: History, Current You cannot boost the growth Evangelist for Google. Practice and the Role of the of grass by pulling at its CPB, Frits Bos, CPB Nether- leaves. It is better to spread Vint Cerf: I have been fol- lands Bureau for Economic lowing SURFnet and the SURF Policy Analysis --

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 14 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 http://papers.ssrn.com/sol3/ Since about 2000, a forward natural gas stock, the fixed papers.cfm?abstract_id=100 looking view on Dutch public capital stock and the financial 0121 finance has become domi- assets. The discounted value nant: Dutch public finance of the natural gas stock was From the Abstract: “Accord- should be sustainable in view 90% GDP in 1970. At pre- ing to the IMF and OECD, the of the net extra costs of ag- sent, it has declined to 20% Dutch fiscal framework is ing and the falling revenues of GDP. p. 34 rather unique, and its design from natural gas.” pp. 9-10 and implementation are Mid 1990’s, Dutch politicians highly recommendable.. This The Economic Structure Im- explicitly addressed the issue paper discusses this frame- provement Fund (FES) was of sustainability by creating work. ... Since 1814, the of- established in 1993. Gov- two funds: the FES-fund and ficial notion of a balanced ernment investments in in- the old age state pensions- budget has changed substan- frastructure had fallen from fund. These should help to tially over time. First, when about 3% GDP in 1970 to ensure sustainability of Dutch debt was excessive, it im- 1.5% GDP in 1993. By ear- public finance in view of the posed the redemption of marking via the FES about exhaustion of natural re- loans. Later, a golden rule of 40% of the natural gas reve- sources and the expected rise finance was introduced, al- nues for financing “additional in old age state pensions due lowing new loans for ‘produc- investments of national sig- to aging. However, the solu- tive’ expenditure. nificance”, the structure of tions offered were only for- the Dutch economy should be mal solutions: 40% of the At the end of the nineteenth improved. ... Since 1993, natural gas revenues was to century, the prominent Dutch the FES has disbursed more be used for financing FES- economist and politician Pier- than 31 billion euro. In the investments. The motto was son stressed that each gen- beginning, the FES- to turn underground assets eration should bear its own investments mainly focused into assets above ground. In burden and should not leave on transport and mobility, particular when cost-benefit excessive debt for the next e.g. roads, railway-tracks and analysis for FES-investments generations. After the Sec- channels. However, now also was not obligatory (see text- ond World War, the classic expenditure on knowledge, box on FES), there was no view on the government was innovation and the environ- guarantee that this results in replaced by a macro- eco- ment are financed via the a higher return than alterna- nomic view: the budget of FES. p. 33 tive options, e.g. extra ex- the state was presented as penditure on education or part of a set of national ac- extra reduction of public counts on the Dutch econ- “Dutch gas re- debt. As a consequence, the omy. Since then, the CPB, serves will be gone FES-fund is important for being an independent insti- changing the composition of tute, provides the official es- in 25 years. public expenditure, but its timates on the macro- contribution to sustainability economic developments. Di- What then?” is not clear. p. 35 rectly after the Second World FRITS BOS War, this new macro-view Under unchanged polices, the aging population will lead to a was combined with a strict The major assets of the budgetary control ... Dutch government are the sharp and structural increase in public expenditure, in par-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 15 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 ticular on state pensions and the Netherlands. Establishing environment designed to en- health care. Government a benchmark ICT technology able scientific and technology revenue from taxes on infrastructure is in that tradi- success. This is the kind of funded pensions will also in- tion. It has given the Nether- economic growth that, with crease, but not enough to lands “the high ground” in pragmatic fiscal policies and cover the extra expenditure this new global environment competent governance, could and the falling revenues from where the world of atoms and eventually replace the income natural gas. As a conse- the world of bits are rapidly generated by millions of quence, in the long run with- becoming one. (Giving the years of geologic history. out policy adjustments public Netherlands “the high debt will explode and Dutch ground” is a nice achieve- The July 2007 study con- public finance will be out of ment, metaphorically speak- cludes with the warning that control.” p. 36 ing, given high ground is given that gas reserves were something the country lacks.) finite (25 year estimated life We end our tour of the Dutch time), it was important to Fiscal Framework paper with The Netherlands is advancing transfer infrastructure sup- this quote: “In about 25 towards gaining “first mover port from FES gas fund basis years, Dutch natural gas re- advantage” in the world of to a structural basis meaning serves are expected to be research and collaboration part of the normal govern- exhausted.” p. 37 networks. ment budget funded by tax- http://en.wikipedia.org/wiki/ payers and the NWO – that is The most surprising aspect of First-mover_advantage . the Dutch Science Founda- the above paper is how tion. The section below, oblivious its author appears called the Dutch Road Map, to be regarding the critical …gaining first provides to find out what role digital technology, and mover advantage happened. especially networked digital technology, plays in redefin- gives the The Evolution ing how the world will work Netherlands the of Knowledge and who will succeed from Infrastructure as a this time forward. The late high ground. Hans Rosenberg understood Generally Accepted this, as does everyone at Creating a world-leading Goal SURF and at other institu- knowledge infrastructure, One of the key questions is tions in the Netherlands ICT that is further empowered by ecosystem. It’s why the gov- how does a physical infra- its open philosophy, is likely ernment in the Netherlands structure become a knowl- to make it more nimble and has chosen to make this in- edge infrastructure? Kees successful than the tendency vestment. Neggers says the process be- of other EU initiatives, which gan in 1994 with documenta- continue to favor large scale The genius of the Dutch nation that formalized the goal. top down projects. It seems tional character is, as we This initiative was refreshed extremely likely that the wrote earlier, the often- and strengthened in 1997 SURFnet7, GigaPort3 and e- demonstrated ability to work and 1998. The Dutch were Science Center initiatives will “leading the duck”, once together to find and execute prosper. They will continue to strategies that win significant again taking the First Mover create a foundation for an competitive advantage for Position, as the EU didn’t be-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 16 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 gin talking about having a tual networks, services and a number of real, active pro- knowledge infrastructure until tools that could be shared grams to facilitate the use of 2000. The EU initiative re- and could attract other pro- the evolving Knowledge In- sulted in the Lisbon Protocols. jects and communities within frastructure in the Nether- broadly accepted national lands. According to Wikipedia, the goals. Lisbon Strategy, also known COOK Report member Ru- as the Lisbon Agenda or Lis- SenterNovem Turning Pol- dolf van der Berg explained bon Process, is an action and icy into Reality what SenterNovem did and development plan for the http://www.senternovem.nl/e how it actually worked. He European Union. Its aim is to nglish/ wrote on the list: The short make the EU "the most dy- answer is that the Ministries namic and competitive The COOK Report is espe- make the policy and Sen- knowledge-based economy in cially interested in SenterNo- terNovem is just the factory the world capable of sustain- vem, an agency of the Dutch that processes all the subsidy able economic growth with Ministry of Economic Affairs. and permit applications that more and better jobs and It’s website states: “We pro- are the result of this policy. greater social cohesion, and mote sustainable development and innovation, both respect for the environment In the Netherlands there is a by 2010".[1] It was set out within the Netherlands and (strict) separation between abroad. We aim to achieve by the European Council in formulating policy and exe- tangible results that have a Lisbon in March 2000. Con- cuting policy. The Ministry of temporary key thinkers on positive effect on the econ- Economic Affairs (COOK Re- omy and on society as a whose works the Lisbon port list member Joost van Strategy is based and/or who whole. Our core competence der Vleuten) is currently is converting government were involved in its creation formulating a policy on how policy into reality. On behalf include Maria João Rodrigues, to get broadband to all. Let's Christopher Freeman, Bengt- of the Dutch government we say for arguments sake that implement policy regarding: Åke Lundvall, Luc Soete, Car- Joost actually got €7 billion to lota Perez, Manuel Castells, --Innovation roll out FTTH to every home. Giovanni Dosi, and Richard --Energy and Climate Change Nelson. Key concepts of the --Environment and Spatial [Editor’s Note: Of course Lisbon Strategy include those Planning “ not likely given the current of the knowledge economy, economic situation.] innovation, techno-economic …if the Netherlands paradigms, technology gov- After this policy is approved ernance, and the "open wants to be truly by the Minister or if it goes method of coordination" beyond her powers, by par- (OMC). competitive, we liament, SenterNovem will http://en.wikipedia.org/wiki/L need to be innovative most likely be contracted to isbon_Strategy. But the Lis- execute the policy (distribute bon Strategy was viewed as a we need to excel… the money in this case).” failure long before 2010. WWW.SENTERNOVEM.NL Van der Berg continues: Meanwhile, the Netherlands It isn’t fancy, but visitors can “SenterNovem will look at the focused on creating a digital quickly click through and find policy as already defined and commons in the form of ac-

Motion Submitted by whereas, the government is gen- OECD are done by the Minis- Marjet van Zuijlen erously investing in physical infra- try. OPTA quite often has an 6 November 1997. structure; input in these areas, though not always. For instance Law- asks the Government to encourage Motion The Parliament, heard the ful Interception or ICANN and that for the destination of the so discussion, the OECD are generally not of called ICES-funds a balanced ap- OPTA’s concern at all. whereas, The Netherlands has the proach is chosen in which there is also scope for strengthening the ambition to be in the front line of COOK Report: An extremely knowledge infrastructure, such as the electronic highway; investments in SURFnet4 and Am- useful SenterNovem page sterdam as an international data that explains the develop- Considering that these development of investment in Knowl- ments until now have seen a slow communications hub, edge infrastructure is found start; and proceeds to the order of the here day http://www.senternovem.nl/ bsik/algemeen/achtergrondin then will work with the Min- specialists in the area fo/index.asp istry in further defining the whereas the ministry consists The website is in Dutch, but nitty gritty of the policy. more of generalists. When it Google translation services That is if the policy says that comes to supporting the provides a rough English the money should go to the Creative Sector in The Neth- translation. consumer and the consumer erlands it acts more as a fa- can then pick a network of cilitator, bringing the right That webside focuses on the choice, SenterNovem will people together. With certain importance of Global Knowl- have to decide whether it is simpler subsidies/permits for edge, identifying it as the key permissable for FTTH- SMEs, it may be just a competitive factor. “The networks to combine the ap- subsidy/permit factory, Dutch economy is increas- plications of thousands of in- where the employees add ingly knowledge intensive. dividuals in a region to get little to the actual policy Nevertheless, our economy in the subsidy directly. Sen- making.” terms of knowledge is not up terNovem will also decide to the demands of our time. how they want to verify The same division can be If the Netherlands wants to whether the money has been seen between the Ministry of be truly competitive, we must used according to the rules Economic Affairs and the be innovative, and dare to etc.” Dutch Telecom Regulator excel. This calls for drastic OPTA. Unlike the FCC or Of- measures. It means that “There is of course not a com, OPTA doesn't write it's more research should now strict line between policy and own policy. OPTA has to exe- focus on innovation and/or implementation... it's a grey cute what the minister has social need. It also forces area at best. Depending on decided. In practice of course more effective deployment of the area in which you are there are many shades of resources and research, for working SenterNovem will be grey there. The Dutch input example by joining forces of more or less involved in the to the EU's Telecoms Frame- the involved parties. The defining of the actual policy. work, E.164 and E.212 num- Dutch government is aware For instance when it comes to bering policy, policies on law- and has taken steps to ex- renewable energy it has ful interception or the inter- ploit innovation opportunities much influence as it employs actions with ICANN and the in the Netherlands in order to

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 18 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 propel it to the top in available for these pro- knowledge infrastructure for Europe.” grams.“ the Netherlands. A 30-year- old history and political sci- Investing Together The vote to fund ence major and Labor Party member named Marjet van a knowledge Zuijlen went to parliament. Senternovem encourages co- http://nl.wikipedia.org/wiki/M investment and public/private infrastructure arjet_van_Zuijlen Previously partnerships, similar to the Ms. Van Zuijlen had worked innovative infrastructure- was virtually several years as a recruiter in sharing partnership recently unanimous… the personnel department of initiated by SURFnet with KPN. But on this day she was Philips. (The Phillips partner- focused on the public inter- ship is discussed later on in SenterNovem’s website pro- est. Her motion asked for the Chapter VIII How A Progres- vides an historical timeline. government to commit to sive ICT Infrastructure Bene- The focus on economic infra- faster implementation of the fits The Economy”.) structure began with creation change towards knowledge in 1993 of the FES fund or infrastructure support, to fi- The SenterNovem website Fonds Economische Struc- nance entities like SURFnet4 explains, “The government tuurversterking, (economic to strengthen the national can not solve everything on structure enhancing fund.) knowledge infrastructure and its own. It is important that And in 1994 with ICES/KIS-1. to position “Amsterdam as an companies, research institu- But the parliament was of the international data communi- tions and governments can opinion that not enough cations hub…” The motion invest in the development, knowledge infrastructure pro- carried almost unanimously. dissemination and utilization jects were granted following It was the turning point that of advanced knowledge. The the 1994 initiation of ICES/ meant projects like GigaPort Interdepartmental Committee KIS1. and VLAM (Virtual e-Science on Enhancing Economic Laboratory Amsterdam) got Structure (ICES) was estab- Then in November, 1997, funded out of FES at the end lished to promote investment something happened that of the nineties. projects that enhance the opened the door to the future economic structure of the for creating a true national Netherlands. . . . . Because of the importance of investing in Since 1993 Five ICES/KIS Calls Have knowledge (KIS) the working Been Issued group ICES/KIS was created. This group focuses specifi- 1994 ICES/KIS-1 for 113 M Euro cally on projects in knowledge infrastructure. Under 1998 ICES/KIS-2 for 211 M Euro the leadership of the group, 2003 ICES/KIS-3 (called Bsik) for 800 M Euro two knowledge infrastructure programs were established in 2008 FES round 279 M Euro 1994 and 1998 under the names ICES/KIS-1 and ICES/ 2009 FES round 500 M Euro KIS-2. There was respectively 113 and 211 million Euros

SenterNovem says this “Netherlands has opted for a - Proposals will then be sent “speed up the 1998 call limited number of strategic to the Committee of Wise which was really focused on knowledge and the pooling of men and to the Netherlands knowledge infrastructure and financial resources and ex- Bureau for Economic Policy resulted in the release of 211 pertise. All under the motto Analysis each of which scores million euro for ICES/KIS-2.” "innovation, focus and them and recommends some [Note the first GigaPort pro- strength. The strategic for funding while turning ject in 1999 was funded out knowledge are chosen based down others. of ICES/KIS-2.] on ideas and needs from the field. . . . . This resulted in Government finally decides As ICES/KIS-2 finished out in five themes: High quality on the winners. The appro- 2002, the umbrella for a space, ICT, Sustainable Inno- priate ministry in conjunction follow-on program was called vation System, Micro and with SenterNovem will then Bsik or Decision on Subsidies nanotechnology and Health, oversee the execution of the for the Investment in knowl- nutrition, gene and biotech- projects. edge. On November 28, nology breakthroughs.” 2003 the government made - As a part of the funding Bob Hertzberger and Kees an amount of 800 million rounds Government may de- Euros available for knowledge Neggers constructed a table cide that only one proposal for this report that illustrates based projects as part of the from a strategic area is also-called "Third boost in- the key events and players in lowed, forcing the area to vestment in knowledge" the FES fund infrastructure really join forces in a collabo- policy process. It can be seen (ICES/KIS-3). The 800 mil- rative way. In the 2009 FES lion comes from the Eco- on the next page. round, for example, ICTRegie nomic Structure Enhancing was asked to prepare a single Fund. - The process starts in the proposal for the ICT area. left upper corner with the “There are seven ministries preparation of a new call. involved in the program, ICES/KIS2 for example is The Dutch ICT Road- namely of Education, Culture such a call. The government map to a Generic and Science, Agriculture, Na- decides the amount of money Knowledge Infrastruc- ture and Food Quality, Trans- available and the rules for ture port, Housing, Spatial Plan- judging the call. These rules ning and Environment, Fi- and processes do not stay The proposal progress mid nance, Economic Affairs and uniform from call to call. 2008 through 2009: Health, Welfare and Sport.” - After Parliament approval, a Since the mid 1990s both the Bsik is to promote coopera- call is published by Sen- Netherlands and the Euro- tion between research and terNovem. Public private pean Union have set up plan- enterprise development and partnerships will then be ning commissions for infra- thus leading to quality net- formed to react to the call structure — and especially for works in the knowledge infra- and proposals written and research infrastructure. The structure. Within these net- vetted within each requisite Dutch Cabinet has the final works, research is conducted industry and agency and its vote on infrastructure pro- that addresses the societal academic counterparts. jects. But the Cabinet must need for knowledge. decide according to a struc-

COOK Report: The procedure for deciding how to use infrastructure funds fascinated me and as I tried to understand how it worked the vision of a kind of ﬂow chart occurred. so thanks to Bob Hertzberger and Kees Neggers for designing the above chart. Follow the arrows. I wish hat the US government would do this. tured and well thought-out (CPB) which renders judg- Science (OCW) to set up the process. ment on the estimated eco- Dutch Committee on a Na- nomic benefits of the propos- tional Roadmap for Large- FES rounds are made on a als. The two findings and scale Research Facilities. This regular basis for funding with recommended actions are became known as the Van specified blocks of gas then sent to the Cabinet for Velzen Commission, named money. They are reviewed by the final funding decision. for its Chair Wim van Velzen. the Commissie van Wijzen (See the diagram above.) Mr van Velzen is a very sen- (the Committee of Wisemen) ior European policy advisor in for technical soundness and The July 2007 CPB Report focusing on technology as- by the Netherlands Bureau cited above caused the Minis- sessment and science. He for Economic Policy Analysis ter of Education, Culture and has also very recently been

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 21 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 named Chair of the Commit- Instruments, com- tion over the next 25 years, tee of Wisemen. He is a how to make the knowledge member of the Christian puting facilities fund infrastructure structur- Democratic Party (CDA). Pre- and storage are ally sustainable — i.e. to viousy he served as Chair- make it supportable by the man of CDA (1987-1994), as becoming generic taxpayer and from regular Senator in the Dutch Parlia- government budgets. The ment (1987-1995), and as a resources. Van Velzen committee and member in the EU parliament SURF ICT Roadmap ICTRegie invited SURFnet, (1994-2004). the VL-e project and other stakeholders to explain what [Editor’s Note: The Euro- they had done in the past, pean Strategy Forum on Re- hoped to do in the future, search Infrastructures (ES- and to estimate how much it FRI) established in April would cost. 2002. The role of the Euro- pean Strategy Forum on Re- Consequently SURF research Infrastructures (ES- sponded on July 8, 2008, FRI) is to support a coherent with a ICT Roadmap proposal approach to policy-making on that described what a Na- research infrastructures in tional ICT Research Infra- Europe, and to act as an in- structure for the Netherlands Wim van Velzen cubator for international ne- would look like. It was pre- gotiations about concrete ini- pared collaboratively with in- tiatives. In particular, ESFRI puts from SURFnet, Nether- Also in 2007 ICTRegie began is preparing a European lands Computing facility responding to a request from Roadmap for new research (NCF), BigGrid (see later in the ministries of OCW and infrastructures of pan- VL-e interview) and VL-e with Economic Affairs for an European interest. the help of others such as evaluation of the entire ICT http://www.lifewatch.eu/inde SARA and Netherlands BioIn- Infrastructure for research in x.php?option=com_content& formatics Center (NBIC). the Netherlands. In an in- view=article&id=5&Itemid=5 Here are some of the data terim report in December points from the SURF re- 2007, the Van Velzen Com- There is increasing broad- sponse. mission had indicated that based interest in many EU high quality infrastructure for countries today in increasing “Modern research is increas- research was an essential the visibility and viability of ingly a global effort, based on prerequisite for hosting any European ideas in technology international cooperation and of the large-scale ESFRI fa- and culture.] resource sharing. Instru- cilities in The Netherlands, or ments, computing facilities to allow Dutch researchers to In responding to Van Velzen and storage are becoming use these facilities elsewhere. and ICTRegie, OCW in effect generic resources, which can ICTRegie stressed the impor- said: “We want you to look be combined through ad- tance of an excellent ICT In- at the long term sustainabil- vanced networks to provide frastructure in a February ity of this infrastructure. We services to an increasingly 2008 report as well. want you to figure out, given diverse research community. the expected gas fund deple-

“The exchange, combination “This proposal concentrates tational science and compu- and integration of data and on creating and maintaining tational life sciences. information has become pos- an advanced ICT research sible as computer and stor- infrastructure in the Nether- “To implement the infrastruc- age facilities have been con- lands. It includes networks, ture we intend to structure all nected via very fast and high computing, vizualisation and necessary activities into a bandwidth networks. The re- storage hardware, as well as coherent program. Such a sulting distributed computer the middleware and generic program will include the fol- systems are developing into a services needed to enable lowing activities: Acquiring, single ICT research infra- modern research. Given the maintaining and operating structure, consisting of hard- importance of international equipment and software li- ware (network, computers cooperation in modern sci- censes; Software hardening and data storage facilities) ence, the infrastructure will and support; R&D to enable and software (web and grid be connected to other initia- efficient usage of the infra- middleware) harnessing the tives worldwide. structure and to facilitate resources. By enriching this new communities.” infrastructure with adequate “The project will build upon tools, which provide services existing facilities and knowl- The proposal to the Roadmap for e-science, a generic in- edge where possible. In par- Committee had to discuss the strumentation for modern ticular, it will build on the re- criteria shown in the follow- research emerges which is far sults from the projects VL-e, ing text box below. In doing more effective than anything BiG Grid, GigaPort and Giga- so they created the inventory the individual research disci- Port Next Generation, NBIC summarized in the in Chapter plines could develop them- and MultimediaN, as well as VI Growing e-science Do- selves. from the NWO-funded remains. search programs for compu-

Funding Criteria And - Science case: potential for scientific breakthroughs; - focus for the Netherlands, either in existing expertise - talent case: potential for brain gain, Netherlands has to or desired expertise; remain an attractive and challenging place to work; - critical mass, availability of sufficient Dutch qualified - partnership: opportunities for collaboration and syn- top-researchers interested in this facility; ergy; - embedding; the facility should be embedded in Dutch - innovation case: interest for society and industry, and international larger research structures or stimulate magnet for new knowledge; further collaborations and concentration of efforts; - business case: financial soundness; - proven will to collaborate, including commitment for - technical case: technical feasibility/challenge. own financial contributions; - connection to societal developments and challenges, like water, energy, health care, security etc.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 23 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 III. The Direction of ICT Infrastructure in the Netherlands in Early 2010 An interview with Wim Liebrand, Kees Neggers and Cees de Laat

The ICT Infrastructure for the Wim Liebrand has been the director of Netherlands Bio- Netherlands continues to Director of the SURF founda- Informatics Center (NBIC). evolve, mature and improve tion since September 2001. As VL-e director he was re- it’s ecosystem. And unlike He explained that SURFnet is sponsible for developing the that classic productivity an important division of SURF functional model that to- graph for Moore’s Law – with for which SURF is responsi- gether with the methodology it’s single curve arching up- ble. There is also a division of multidisciplinary collabora- wards – a collaborative, that deals with software li- tion form the basis of the network-based, re-usable in- censing for staff and students project. frastructure evolves, matures in research and higher educa- and improves in many direction institutes and next year Kees Neggers is one of the tions at once. SURF will start a division that founders of SURFnet in the deals with shared services for Netherlands and has been a On Tuesday, the 17th of No- students’ admissions and en- Managing Director there since vember, 2009, many of the rollments in Higher Educa- 1988. Since the mid-eighties leading figures in the ICT In- tion. After the implementa- Neggers has been on the frastructure for research in tion of the ICTRegie advice forefront of both the technical the Netherlands sat down we will also be responsible for and organizational evolution with The COOK Report at the the computing and storage of the internet and research Supercomputing 09 meeting facilities and will be support- networking developments in Portland, Oregon, to pro- ing a new e-Science Research worldwide and involved as vide a progress report on the Institute. initiator and Board member state of Netherland’s infra- in many international net- structure as a new decade Hans Dijkman is responsible working related organizations begins. for E-science at the Univer- and initiatives like RARE, sity of Amsterdam. He has a The participants included: PHD in Chemistry and an MBA. He has 20 years experience as ICT manager of ICT departments from research and education institutions. His Management philosophy: Mathieu Weggeman, Òn management of profession- als? Don't!

Bob Hertzberger is director of the VL-e project as well as one of the directors of Big Wim Liebrand, Director of SURF Grid and till recently adjunct Hans Dijkman, Director e-Science University of Amsterdam

TERENA, Ebone, Internet So- of the Spanish Armada by tation plan for it. ciety, RIPE NCC and most making it possible to build recently GLIF, the Global more and cheaper ships. Liebrand: When you talked Lambda Integrated Facility. to Kees Neggers a year ago, you have built the the ICTRegie Report was still Cees de Laat is professor under construction. It took a and chair of the System and 21st century couple of months of review Network Engineering group at before it was fully supported the University of Amsterdam. infrastructure by the government — that is, Research in his group in- the Ministry of Education and cludes optical /switched It is impressive that the the Ministry of Economic Af- Internet for data-intensive Netherlands managed to fairs. So they together - re- TeraScale e-science applica- gather together the neces- search and industry and edu- tions, Authorization, Security sary academic, political, sci- cation –supported the ICTRe- and Privacy in distributed entific and research funding gie document. They asked systems and Semantic web to entities in the Netherlands — SURF to work with these enti- describe e-Infrastructure. He to produce what appears to ties to build a transparent serves as board member of be a unique approach to the plan for an eco-system for ISOC.nl, chairs GridForum.nl, careful coordination and technology transfer from ba- is co-founder of CineGrid.org planning with maximum cost- sic research. and the Global Lambda Inte- effectiveness of public and grated Facility (GLIF). private funds — resulting in The total financial picture has http://ext.delaat.net/ leading-edge research and not yet been realized as the technology transfer. In many result of the impact of the The Technology Re- respects, you now have the global economic crisis, but most advanced optical net- search Ecosystem of the good thing is that every work in the world. one is collaborating and we The Netherlands are finding piece-by-piece So where do you go from financial support with which COOK Report: No other here? How did you get to this to implement the plan by country has put together point? What have you ac- means of cooperation be- such an extraordinary eco- complished and where are tween industry government system of optical infrastruc- you are headed in the near- and the research institutes. ture and leading-edge re- to-midterm future? search and industrial tech- [Editor’s Note: The recom- nology transfer applications Neggers: In December, mended yearly budget for the as has the Netherlands. This 2008, the ICTRegie published implementation of the IC- is a true 21st century infra- its recommendations for the TRegie advice is 62.5 million structure that is, in a way, future of the national ICT in- Euro, split over the various equivalent to the remarkable frastructure. The government infrastructure items as 17th century infrastructure of took that advice into consid- documented in chapter VIII, windmills and canals. That eration, and by May, 2009, page 26 of the ICTRegie ad- was a cooperative pragmatic they had decided the ICTRe- visory from December 2008. environment that enabled the gie advice was good and The Networking component invention of the mechanical wanted it to be followed up. of this is 10 million Euro per sawmill in the 17th century. Consequently they asked year. And it is already se- In turn that led to the defeat SURF to make an implemen-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 25 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 cured for 5 years for Giga- network is a condition sine useless. Port3.] qua non in all scenario’s. COOK Report: So when you COOK Report: Is it the Collaboration and the say software, you are talking shared vision of all partici- E-science Paradigm about “middleware”? pants that has provided the critical difference in the De Laat: Yes, the generic COOK Report: Are you able Netherlands? to talk about the plans? part and the application domain specific part - both are middleware. ...the network is Liebrand: The plans themselves are described in the Liebrand: E-science has al- a condition advisory report. And to- ways been difficult to label, gether with the Dutch NSF sine qua non… but we understand it in the (which we call NWO) we are broad definition provided by working to build an e-science Neggers: Yes I think you Bob Hertzberger later in your research platform… which is a can say there is among the report. However, e-science is combination of fundamental stakeholders a shared vision also related to computational generic research, combined that was achieved by exten- science, to simulation, and to with the infrastructure. That sive consultation among the grid computing and visu- in turn is combined with the them before the report was alization infrastructure. Take network. published. Consequently we all of these together and you have a consensus among the will get what we consider to stakeholders that this was SURF be the e-science paradigm. indeed good advice. Now at You will be that point the government De Laat: It is a new way of needed to take it on board the umbrella doing science. and make a financial com- organization to mitment to implement it. COOK Report: It is the dikes govern this whole and canals of the twenty first As Wim has pointed out the century. government said: SURF – eco-system… you will be the umbrella or- Dijkman: Not just the dikes ganization to govern this and canals. It is also the COOK Report: Is this what whole eco-system. Now give ships. us a detailed plan as to how Bob Hertzberger is doing? you will do this. And until we De Laat: Of course, you also De Laat: Y e s , B o b have seen your plan and de- need the correct people to Hertzberger is talking about cide that it is a good and steer those ships. It needs the software on top of the workable plan, we won’t knowledge and tools by infrastructure. The under- commit the money. This is which the infrastructure can standing in the Netherlands is the position we are in at the be used. moment. The financing of the that software is actually infrastructure. Without the network innovation however Software is way more expen- middleware to enable the in- is already committed. Every- sive than hardware. For frastructure for applications one was convinced that a hardware you can get quite a the infrastructure itself is discount for an expensive

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 26 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 cluster or supercomputer. we have to be careful that found out, we published our But when, for example, you what we build works… and results in scientific and engi- consider the man-hours that not just inside the Nether- neering journals. It was, and have gone into creating lands. It has to work often still is, not trivial. Globus, (a particular imple- throughout the global re- mentation of the Grid mid- search community. That is COOK Report: Was the dleware layer) you under- one reason we have all come amount of data so huge that stand that a floppy with a here to Portland, to make your buffers filled up while Globus installation is very sure what we are building is waiting for an Ack? expensive. The real cost for available globally. The same creating the correct software is true for this grid and e- De Laat: Yes – we had to is typically underestimated. science middleware layer. No re-engineer the transport single institute – not even a protocol stack. We get repe- Globus is open source. Its smart guy like Cees de Laat titions of all these kinds of libraries can be freely down- can do it on his own. problems as we increase our loaded. But the manpower bandwidth (1 - 10 - 40 - 100 required to manage and This is why the government Gbit/s). I can point out some maintain these generic envi- of the Netherlands is inter- rather common problems in ronments of the middleware, ested. Because they need to networks which pop up again which make the grid and provide the lubrication that and again every two or three clouds possible, is always keeps things going forward. years. For example, in some under-estimated. And the cases UDP traffic stops work- middleware must be kept up COOK Report: What is the ing after some time on layer to date so that the biologist, role of GLIF in all this? 2 engineered lightpaths. for example, can use the grid Why? Because UDP is unidi- without having to become a Neggers: We started GLIF rectional, and your Ethernet computer scientist. because… we were aware switches forget where the from the start… that if we destination is. And they start The real cost built something that works flooding. between Utrecht and Amster- for creating the dam, but is not connected to We had to reverse engineer correct software the rest of the world in a vi- the hardware in the link to able fashion, it is really use- Chicago. With lambdas and is typically less. This is why we peered lightpaths, you really open up with Tom DeFanti and Maxine the infrastructure to the ap- underestimated. Brown at STARTAP, and then plications, which become StarLight in Chicago, from very sensitive to the proper- Neggers: This is very simi- day one in 2002. At that ties of the hardware if you lar to the physical network moment it achieved a critical remove the router layer. We itself. No single institute mass globally. could really measure those could build the network to properties, the buffers and provide necessary connec- De Laat: Of course that first bottlenecks and so on… and tivity. This is why you need lambda connection between given the round trip time, we cooperation and investments Amsterdam and Chicago could predict how certain up front to create the infra- didn’t work just like that. We links would perform with cer- structure. We have done this had to research why it did tain protocol stacks and ap- for the network. And then not perform and when we plications. I had to go back

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 27 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 into Cisco… and up to esting. They said, you are that even when the box is Chambers-minus-one… to get using our equipment incor- used in the metro, but serv- their attention to the limita- rectly. This is a metro box, ices some connections that tions in the hardware. and you are using it wide go on long distances, the area. problematic effects would still Neggers: The initial re- be there. sponse from Cisco was inter- De Laat: Then we proved COOK Report: In 2002 you were addressing the recurring problems of the network and the middleware. Did you find a similar set that affected the high performance computers?

Liebrand: Yes. But viewed from a slightly different perspective. There are so many cores – as in multi-core processor -- that we have to think about how to use them properly. The old ways won’t work anymore.

Return on Investment in the High Performance Infrastructure

De Laat: There are indeed similar issues in high performance computers. There is an accretion of knowledge that is built up, and it needs to be maintained as the infrastructure grows. One of the tasks for SURFnet and for e-science is that the knowledge base gained from using this extreme infrastructure must be maintained.

COOK Report: What do you mean when you talk about At the Netherlands booth SURFnet was demonstrating how easy it is to upgrade from 10 to 100Gbit/s second in its existing Nortel Common Photonic SURF providing the “oil” that Layer equipment. Two SURFnet6 nodes were installed interconnected by an lubricates this ecosystem? artificially long fiber of 1120 km. See the picture on p. 37 below for a close up of the fiber spools seven of which were linked together.

Liebrand: Actually it is a industry, and sometimes rather unique phenomenon from participating partners. Neggers: It is interesting to because the cooperation And then you give that to the note that the Government within our “umbrella envi- users as a production infra- innovation grant which en- ronment” has existed now for structure. Meanwhile, as the abled us to build the first some 25 years. We began in users are gaining experience, S U R F n e t n e t w o r k w a s 1986 with SURFnet version you start with a second round roughly the same amount as 1.0 at 9.6 kilobits per- of innovation… that builds an the grant that enabled SURF- second. even more advanced infra- net5 with 10 gig lambdas and structure. We innovate suc- SURFnet6 as a full hybrid op- The collaboration model is to cessive networks. So with tical and packet switching build a pretty advanced infra- SURFnet6 mature, we are network. structure, with innovation getting ready to launch money from government and SURFnet7 soon. Liebrand: More bang for

This picture shows a close up of one of the OME6500 systems in the racks at the booth. The only thing needed is two free slots in an OME6500 to put in a 100Gbit/s transponder. The 100G links operate on the existing 10 Gbit/s optical wavelengths. In July 2009 Nortel and SURFnet successfully demonstrated that this works in its live network over a distance of 1244 kilometers on its link between Amsterdam and Hamburg. More information can be found on: http://sc09.delaat.net/ http://www2.nortel.com/go/news_detail.jsp?cat_id=-8055&oid=100257448&locale=en-US&lcid=-1 If you are interested in a YouTube movie of the demo go to: http://www.youtube.com/watch?v=P5jVGpjbdkQ For a more technical demo in the Nortel labs go to: http://www.youtube.com/watch?v=zEUDiRWNmII&feature=related

Netherlands Bureau ceived supporting letters for Economic policy from stakeholders outside the Analysis (CPB). The direct Research and Educa- government re- tion Community — who al- quested this as part ready had seen the positive of the policy proce- effects of our activities for dure for proposals society as a whole. This, to- funded under our gether with a further expla- knowledge infra- nation from our team, con- structure funds (as vinced the CPB economists Close-up of the fiber spools. There were 7 spools of 80 km in one direction for a total of 560km described on pages that what we did was impor- with a loopback of the 100G wavelength for an 14 to 19 above). tant. And they endorsed our additional 560 km of fiber for a total of 1120km between the two optical nodes for the 100G demo They did econometric plans again. Also the soon the Netherlands booth at SC09 in Portland. studies beginning at called Commissie van Wijzen, The nodes were less than 2 meters apart on the show floor shown in the picture on page 35 above the end of the 90s. “The Wisemen”, which at that In the 90s invest- time was the final advisor to ment in these tech- the Government, also recog- your buck. This is Moore’s nologies was regarded as a nized once more the impor- Law applied to innovation no-brainer. But after the dot tance of our efforts for the networks. com crash, it was no longer entire economy. so easy to defend invest- De Laat: Just for fun I have ments. calculated very roughly the NetherLight and increase in capacity for the We had a difficult time con- AMS-IX are both Netherlands from SURFnet4 vincing them to finance our real examples of through SURFnet6. I multi- innovation again out of these plied the number of kilome- infrastructure funds. Initially unanticipated ters, first of copper, then of they saw no need for a new glass, times the capacity. investment in research net- payoffs from the The growth over 14 years is working or for tapping these SURFnet projects. about 10 orders of magni- special funds again. By that tude. Actually, these in- point they considered the In 2008 things became a bit creases are much faster than network to be nothing more easier again. Just before the Moore’s law! than a normal tool for the government evaluation of our researchers. Consequently, it SURFnet7 and GigaPort3 Liebrand: It is difficult to do should be paid for only by the plans, an evaluation by an a precise calculation, but for government entities that independent party was pub- each Euro invested, the pay- supported research. Their lished. It focused on the back to the economy is be- initial conclusion was that whole ICES/KIS-2 program tween four and five Euros. these entities should pay for that financed the first Giga- This is my gut level guessti- the network from their regu- Port project during which we mate. Maybe Kees knows lar budgets. They no longer completed SURFnet5. This more advanced models? saw investing in the network was five years after the end as infrastructure as having a of that program and the ef- Neggers: For SURFNet5, 6, broader national impact. and 7 our plans were evalu- fect on the economy of infrastructure components like ated by economists from the We then asked for and re-

GigaPort had become very De Laat: And because of that. visible. This report appeared this, (the grid infrastructure a few months before we had and the knowledge we have,) COOK Report: How will to defend the SURFnet7 pro- EGI.org (the European Grid Nortel’s bankruptcy affect ject. So we ended up receiv- initiative) has decided Am- support for that technology? ing a very positive endorse- sterdam is the place to be. ment for the GigaPort3/ Neggers: We are not wor- SURFnet7 project, both from Liebrand: It is difficult to ried. The moment it was an- the Netherlands Bureau for calculate the exact payback. nounced, big competitors Economic Policy Analysis and But if you analyze it the way knocked on our door and the Commissie van Wijzen. utilities calculate return on said, “Don’t worry. We will be The Commissie van Wijzen investment, then I would say able to maintain your net- recognised at an early stage there is a factor of four or work instantly should Nortel the importance of the inte- five. ever actually fail.” There is gration of our different ICT such a large installed base Infrastructure components… Neggers: The point, of that there is a solid market including the e-science com- course, is the spill over ef- for maintaining this equip- ponents. So this provided fect. The entity that does the ment. We believe there is momentum for further gov- investment will not see the little risk to the network, and ernmental support. returns coming back directly we believe the technology is to it. They materialize else- so valuable it will emerge in- Liebrand: One of the out- where in the economy. Be- tact from any take over. comes that we never ex- cause this is infrastructure, it [Note: Ciena recently an- pected was the pre-eminence affects all sectors. This is why nounced that it has signed of the Amsterdam Internet government must play a part agreements to purchase opti- Exchange, the AMS-IX. AMS- as the investing agency that cal and Carrier Ethernet as- IX and NetherLight are both best crosses sectors. sets of Nortel’s Metro Ether- examples of unanticipated net Networks (MEN) busi- payoffs from the SURFnet SURFnet7 Builds on ness. Continuation of Nortel’s projects. Nortel Common photonic product line seems assured.] COOK Report: Did AMS-IX Photonic Layer, Next (along with all the other basic Generation Ethernet SURFnet6 was composed of infrastructure) help Dirk van and E-science five optical rings connected der Woude and Herman Middleware by optical electrical optical Wagter get the Amsterdam converters. We are removing city fiber build off the ground COOK Report: I understand these converters and will re- including getting EU ap- that SURFnet7 is funded, but place them by optical proval? what exactly is it? switches. This will make all of the Netherlands a single opti- Liebrand: Yes. I fully agree Neggers: SURFnet7 builds cal network — one no longer with your conclusion. The on the foundation of the hy- dependent on any kind of re- technology position of Am- brid network of SURFnet6. generation. This will permit sterdam is directly due to the The Nortel common photonic instantaneous point-to-point investment in the networking layer is still state of the art. connections anywhere in the infrastructure. Therefore, we will build on Netherlands. Today a circuit switched light path from the

Year 2009 2010 2011 2012 2013 Work Package  Flexible  40G waves introduced  New  Technology photonics  100G tested photonics Assessment introduced features 1. Photonics  New Planning tool  Architecture into the introduced Study into (PHO)  Study new network next gen technologies photonics for SURFnet8  Technology  Procurement process  SURFnet7  New NGE  Evolution Assessment SURFnet7 NGE features 2. Next into services up introduced Generation  Architecture  Proof of Concept tests next steps Ethernet study and preparations for and  Design for (NGE) new service running SURFnet8 provisioning  Multi-domain  Federated and Multi-  Proof of  e-Science dynamic domain lightpath Concept for Services 3. Enabling lightpaths Services introduced e-Science available Dynamic Proof of  Integrated resource Service Services Concept (EDS) services, studies and  Architecture Proof of Concepts Study  NetherLight  NGE services available  NetherLight  New NGE upgraded to at NetherLight ready as features 4. Nether- scalable new  Additional 10G “dynamic introduced in Light and platform GLIF Open NetherLight global con- intercon-tinental lambdas available Lightpath  40G nectivity Exchange (INT) intercontine ntal lambdas available  Technology and legal  Wireless  Wireless  Hetero- 5. Mobility Assessments application technology geneous and Fixed-  Operator & Vendor Living lab and mobile service Wireless scans established application offering (MOB) with Pilots with operators operators 6. IP Innova-  Core router  IPv6 support for  40G or  Architecture tion (FIP) upgrade connected institutions 100G core Study future trunks of IP routed  Testing with 100G introduced networks interface  Network Operations Center renewed 7. User Par-  EYR Lightpath  EYR 3 preparations  EYR 3  EYR 4 prep.  EYR 4 contest ticipation & contest  3-5 new research- contest  3-5 new  GP3 closing Knowledge  Contact with projects  3-5 new research- event Dissemina- research-  Knowledge domains research- projects  Knowledge tion (DIS) projects  Events projects  Knowledge domains  Knowledge domains domains  Events  Events

GigaPort3’s main deliverables over the years, highlighting 2010.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 32 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 south of the Netherlands to – that is, in large scale. The where we want to go with the north must pass through major purpose of next gen- next generation Ethernet. But three optical rings, with OEO eration Ethernet is achieving more studies are necessary. regeneration at each ring this scalability. transition. SURFnet7 means COOK Report: Will advanced this is no longer needed. We Local networks are Ethernet networks like SURFnet7 lev- will also optimize the topol- and will stay Ethernet. Eth- erage enable reducing elec- ogy to make it even easier to ernet connects everything. tricity use and CO2 emissions go from a to b anywhere in That means we will have a by avoiding optical-electrical the Netherlands. very user-friendly network –optical conversion? layer at layer two. Of course We will move these lightpaths Ethernet is also much Liebrand: Yes. In fact, Bill to a more dynamic user con- cheaper than SONET/SDH St Arnaud is likely working on trolled interface. Using a GUI equipment. Currently the op- metrics about this. He wants web interface, you will be tical, electrical optical con- everyone to measure their able to decide where in the version in the photonic layer footprint and try to improve it Netherlands you connect your is SONET/SDH based. SURF- in the next couple of years. light path. And we hope to net7 hopes to get rid of extend this globally via the SONET/SDH entirely so that [Editor’s Note: On Decem- GLIF. in the end it is just photonics ber 8, 2009 Bill St.Arnaud and Ethernet if we are suc- replied to my query: No hard This is the Layer One optical cessful and the equipment data yet. But the savings will part. On top of this we are market supplies it. be relatively small in going planning to introduce a next from OEO [Optical to Electri- generation Ethernet service Liebrand: This will take an- cal to Optical] to OOO [all as Layer Two. We call it other four years. optical] with no electrical re- “next generation Ethernet” generation. Routers are the because details of the new Neggers: Yes. huge energy hogs and a OOO Ethernet standards are still network significantly reduces in flux. De Laat: This is also a re- number of distributed routers search question, because and allows you to locate core COOK Report: What are the SONET/SDH uses TDM and routers at a site that uses parameters that will define gives you deterministic be- renewable power. SURFnet the next generation of Ether- havior. In Ethernet you de- already realized this reduc- net? pend on certain protocols to tion of backbone routers in separate traffic and the leaky 2005 in the hybrid SURFnet6 Neggers: Actually it is fairly bucket mechanism to shape network.] simple. It must be a scalable and police each flow depend- service. Ethernet today, in ing on the QoS parameters De Laat: Here are some the sense of circuits, is not you set. We did experiments rough metrics. Each Fiber scalable. You have a limited with Nortel last year. Our takes 80 colors and for each number of VLANs, and glob- experiments showed video color you need a separate ally you will not be able to traffic could be run in such a laser taking between 25 and build any really useful net- network if protected from 35 Watts. A complete rack work out of it. Even the high amounts of “noise” traf- would take 80 times 30 or Netherlands is too big for this fic. This seems to be a sig- 2400 Watts. If you use if you want to use it all over nificant step in confirming higher bandwidth you need

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 33 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 some more Watts. The state and back a distance of more the technology for a while. of the art photonics is such than 1244 kilometers without It’s good to notice that there that you can reach much fur- regeneration. From Amster- are some policy problems ther distances nowadays dam to Geneva we have our that may impact on what the without regeneration and own dark fiber that we will technology can do. have much better ways of light with 40 gigabit/s correcting signal distortions lambdas from day one with COOK Report: I conclude caused by intermediate fibers no regeneration the whole that if you can do this, then and photonic devices. These way from Amsterdam to Ge- you have an infrastructure to new technologies of the last 3 neva. You see therefore why which you can connect people years are now coming into we are not yet ready to sacri- in whatever way and by products that make it possi- fice the Nortel photonic net- whatever means. Then you ble to use wavelength selec- work. have an infrastructure that tive switches [WSS]. Such a renders what belongs cur- switch contains a small cop- De Laat: You should also rently to the incumbents ab- per box that takes 9 fibers as note the impact of this on solutely meaningless? inputs. The box is capable of European networking where selecting colors out of any it becomes possible to cross Neggers: Yes. But let’s go fiber and inserting the mix of national boundaries photoni- in that direction more slowly. colors into an output fiber. cally instead electrically. First of all I want to make it And this copper box takes not simple to connect and this is more than 10 watts to oper- The point is that you are now not the case today. There- ate. able to extend the model of fore an element of the SURF- photonic communication to net7 plan is called Enabling Instead of 70 to 100 watts all countries. You can reach Dynamic Services where we per lambda, a switch contain- 1000 kilometers the distance want to make it simple for ing WSS’es can switch a few from Amsterdam to Geneva researchers to not just use 100 lambas while using only which is about half of Europe the network but also to use 100 Watts for the photonic without OEO conversion. You resources available via the devices and controller. With can also build Lambda Ex- network in a very simple co- this box you don’t have to changes. herent way. have a receiving and sending transponder for each color COOK Report: Nether- In short I am talking about pair. You eventually need a Lights? [See page 9 above.] the integration aspect where transponder, but if you have we have the data, the super- the correct one, you can po- De Laat: Lower layer Neth- computing and the e-science sition them at the edge of the erLights. middleware to glue this all network. together. This middleware cross national layer is also a main part of Neggers: But getting back SURFnet7 development. to the SURFnet7 plans – we boundaries will have higher speeds of photonically You have better and much course. One hundred faster photonics. Then you gigabit/s will become the instead electrically have Ethernet that makes it norm. We have already easier to connect and switch. demonstrated 100 gigabit/s Then you have the middle- Liebrand: But let’s stay with from Amsterdam to Hamburg ware layer that makes it easy

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 34 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 to bring in all kinds of re- they will be left out until they tion community with capa- sources and allow research- slip the “incumbent boat an- bilities ahead of what will be- ers and people to collaborate chors” off the ankles of their come commercially in the on shared problems. economies? end.

Liebrand: It is useful to see Neggers: Right. COOK Report: When you this in context. This is a talk about GigaPort3 it re- world class network in a Liebrand: No, I am sorry. minds me of Nico Baken’s small country, the Nether- It is more difficult. It is like concerns. Where does Baken lands, 200 by 300 kilometers the prisoner’s dilemma. It is fit into all this? and the moment you pass in our own interest not to the borders you have also succeed too well, but to Liebrand: Nico Baken par- some political problems. It’s share the profit with the oth- ticipated in the SURF founda- desirable to notice that there ers, and to collaborate to- tion as a member of the Sci- are some policy problems gether. entific Council. His schedule that may impair the full po- is such that he can no longer tential of technology. COOK Report: Collaborate afford to spend a large with your sister research amount of time with SURF, Neggers: Gordon is well networks in other European but he is well aligned with aware that there are people countries? our goals. As far as KPN in Europe and elsewhere who goes, there is a big difference will want to slow you down if Liebrand: Yes. between KPN’s research and you go too fast. These are development and KPN’s some of the headwinds we De Laat: Science is a col- commercial part. face in our march to the fu- laborative effort that crosses ture. boundaries. Neggers: KPN does very good R&D, but from our point Science is a GigaPort3 – E-science of view they are slower to to E-life bring the results of that R&D collaborative into commercial networks than we would like to see. Neggers: Referring to the effort that crosses We think KPN will be very “netness issue” that we dis- interested in this wired wire- boundaries. cussed earlier at the Portland less interworking. Why? Be- show with its author Sheldon cause we at SURFnet are not COOK Report: Let’s assume Renan, one of the new ele- proposing to blanket the you do this and it gets the ments in the Gigaport3 pro- Netherlands with new radios. expected and hoped for re- ject is to have full mobility But we want to make use of sults for the Dutch economy, and a seamless integration of the investments of public op- because the infrastructure next generation wireless with erators, and develop coop- means being able to do our fixed infrastructure. This eration between public and things in the Netherlands that is a longer-term issue be- private investment. There can’t be done elsewhere. If it cause the frequencies have are enormous cooperative makes you bloom, thrive and not yet been auctioned. But possibilities. grow, won’t Germany, France we want to be there and try and the UK have to take no- to create a private network De Laat: Certainly true for tice and begin to understand for our research and educa- the sensor grid. The pro-

grammable controllable network that I showed you last year is part of the strategy to be able to influence and control the infrastructure. Last year we demonstrated programmable sensor networks for monitoring dikes (see: http://sc08.delaat.net/). With wireless many more sensor networks are imaginable.

applications and networks need to be aware of

The Dutch Exhibit at portland Supercomputing 09. Unfortunately shot each other after closing. On Cees de Laat's website http://sc09.delaat.net/ you can ﬁnd all the details of the SC09 Dutch booth and its demo's. The network itself is not sufficient anymore. It needs to

reach out to the applications attributes in ways that the and to the environment. The semantic web can be applied COOK Report: When the applications and the networks to the whole as an interlinked semantic web convergence need to be aware of each fabric. Making the relation- with the service web is com- other. A lot of the work that ship between the networks plete then these translations we do in describing infra- and everything connected to will become almost auto- structure and applications them is very important. matic?

De Laat: Yes - for example I can show that there is a lot of semantics research going on for applications, e.g. in biology DNA sequencing. There is also a layer of semantics in media content. And if we can link these semantics to infrastructure you get location aware content and location aware applications. And then you can really think about how to optimize the work that you want to do.

COOK Report: Is what you are doing in e-science intended to transition into what you might call e-life? Will the tools and infrastructures that you are building now be tran- sitioned to support the greater virtualization of everyday infrastructures?

Neggers: Of course.

COOK Report: Isn’t that a very deep and important part of what you are doing?

Neggers: This is exactly why the government and in particular the ministry of Some tile displays at the Holland booth in Portland. These are being fed from economic affairs is supporting Sara computers to the booth on the show ﬂoor in Portland. The top tiles are dis- us with additional funds and play of changing ocean temperatures with the continents in white. The canal and birds and four times HD being streamed more than ﬁve thousand miles. are not just leaving us to the support of the ministry of re-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 37 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 search and education. It is Sheldon’s example of the my house and involved my the spill over effect to society bathtub that catches the per- neighbor as I found out when as a whole that justifies this son when he falls. I was digging into application broader support. of wi-fi in education 14 years Neggers: It is this kind of ago. De Laat: Definitely the next lateral transectoral thinking step after virtualization and that we want to trigger with Neggers: Lucent at that new methods of e-science is Enabling Dynamic Services. time had three labs in the the virtualization and new It is the horizontal infrastruc- Netherlands. methods of e-life in the ture that glues all this to- community. It will take an- gether. The Sorcererʼs other five to ten years to get Apprentice Race there but, as I also tried to COOK Report: Is the going point out yesterday, life as is wireless component part of Between the Creation currently lived on earth is not SURFnet7 or will it be a part of More Bits and the sustainable for ten billion of SURFnet8? Means of Storing people. Them Neggers: In the GigaPort3 Neggers: As a very straight- plan we will investigate this COOK Report: Let’s explore forward example, Philips is together with industry. e-life more. I remember that slowly dropping out of con- SURFnet8 likely will bring the Sheldon Renan recounting a sumer electronics and mov- integration of what we learn talk John Seeley Brown gave ing into health care. And for with GigaPort3 into the net- in 1994 where he explained them health care means a work fabric that we are build- that typically you cannot sus- connected person. Wherever ing. We will ask government tain productivity increases he is and that is why you again for support to make beyond 10-12% per annum need the seamless interwork- this integration happen in in the world of atoms. But in ing of wireless and wired to SURFnet8. the digital world - if you make that all happen. For combine the annual im- them this is their core busi- COOK Report: And a signifi- provements in productivity of ness. If the infrastructure is cant reason for on going sup- both hardware and software - not there, new tools to support is to keep the Nether- you may be able to achieve port better health care will lands competitive with the and sustain productivity in- not be there either. larger countries? creases of 100% per annum. The way John expressed it, De Laat: “Our monitoring Liebrand: Yes. Absolutely. although Moore's Law is fo- show that you are due for a cused on hardware (semi- heart attack in five minutes, COOK Report: This is the conductors), improvements please come to the hospital.” same path that the Danish in software appear to be government has followed more significant. COOK Report: Sheldon Re- very productively. nan told me that he has had Software is much faster than a lot of discussions about this Neggers: Did you know that hardware. New software run- with Jacob Bardram who is wi-fi is a Dutch invention? ning on old hardware will out- the head of pervasive com- perform old software running puting for healthcare in De Laat: In fact it was in- on new hardware. Experi- Denmark. And we also have vented two kilometers from

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 38 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 ments at Xerox PARC con- the generic tools that allow cluded that anything you the data of different disci- The film industry would like could do digitally would be plines to be combined to en- to store everything that much more successful than able the next steps of further comes out of Hollywood for things done in analogue fash- progress. 200 years. They are moving ion. from 4k to 8k and now every- If you can’t do that and your thing is done digitally. They Both Sheldon and JSB are competitors can, then you also want to store the work- saying that Moore's Law is no loose. flow, everything involved in longer the most important making each movie. Storage source of new opportunity. Neggers: But there is also needed per movie will be be- JSB said software improve- another challenge for the tween one and three peta- ments were more significant world of bits. In the data en- bytes. They produce 600 than hardware improve- vironment, the creation of movies per year that may ments. And Sheldon is saying bits grows like mushrooms make it to academy award that connectivity improve- because bits are so easy to (Oscar) consideration process. ments are more significant create that like the sorcerer’s They want to keep all 600 be- than hardware improve- apprentice it becomes diffi- cause they have no way of ments. Sheldon also points cult to put any limits on their knowing today what will be the out that as you reach a situa- creation. Storage of them cult movie 50 years from now. tion in the real world where takes a lot of resources. The you’re running out of re- question becomes whether If you look at these data resources, you are forced into the increase in storage ca- quirements, the requirements the digital world. That is pacity is outpacing the proof the Large Hadron Collider your only hope apart from a duction of bits. are mere noise. You must Malthusian reduction in popu- look at the data evolution lation. COOK Report: And doesn’t curve. People experience 2k connectivity the question become one of and then 4k and then they whether the environmental want 8k. improvements are impact of electricity and cool- ing needed is sustainable? The general lesson for this dis- more significant cussion is that as the infra- than hardware De Laat: There is another structure grows faster and issue. While you may even- more capable, scientists will improvements tually get the capacity of the find new problems that will library of congress into a sin- continue to stress the infra- [Editor’s Note: –in the gle disk drive, you will find structure no matter how fast it spring I will publish an in that scientists in massive grows. depth exploration of netness data producing experiments (also known as Renan’s law) as in LHC tune the amount of COOK Report: I sense that with interviews done in Port- data they keep. As the ca- there is an evolutionary proc- land, November 17th and 19] pacity of available storage ess for conceptualizing what increases you will find that you are doing. Would you De Laat: The same holds they measure way more vari- elaborate further? true with e-science as it be- ables because keeping more comes a global and multidis- data allows them to do more Neggers: Well as Wim says ciplinary effort as you need sophisticated analysis. the government gives you an

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 39 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 innovation budget but at the cal and research community the pragmatism? same time the users are and the end user community. Liebrand: It is true we have committed to pay for the use We are trying to interconnect to cooperate and be competi- of the new infrastructure in them. tive. But also can’t forget the the end and, therefore, they role of the pirates. In other are also deeply involved in the Neggers: Some people say words, if you excel at building process of developing it. We that because our citizens have a world class infrastructure, are not a laboratory that has a been forced to work together others will become jealous. few techies on board and tries to protect themselves from the And when jealousy rules, col- to do nice technical things and rivers – actually floods from laboration fails. By some de- then hope that they will get the river are more serious a gree of trial and error we have used. To the contrary we op- threat than the sea – the abil- discovered you need to excel erate with constant feedback ity to have a common respon- in certain fields, but also be from our user community as sibility for the greater welfare willing to share with your represented by Wim Liebrand. is ingrained. competitors. SURF is managed by the users and not by the government. SURF is managed COOK Report: And if you do The government trusts the us- not do this? ers to integrate the infrastruc- by the users and ture and not the ministry. The not by the Liebrand: In the long run you users are in charge of their will loose the game. It is work. government. again basically the prisoner’s dilemma. Liebrand: SURF embraces COOK Report: A source of two communities: the techni- Neggers: This is why we are so active in global collaboration. Because it would be useless to have a nice network that connects everyone inside the Netherlands and nowhere else.

COOK Report: What you are doing is so powerful. What I find ironic is that the power is not well recognized, nor is it understood. People tend to think in terms of their own specialties. No one does a lot of connecting of the dots. But as you saw, when I introduced you to Sheldon Renan, his focus on the interconnectedness Left to Right: Sheldon Renan, Kees Neggers and Cees de Laat at the Dutch Booth. Sheldon is giving an informal exposition of Renan’s Law - Everything of everything may begin to wants to be connected to the “ﬂying” Dutchmen. All parties agree that these help people see the larger are important ideas -detailed exposition coming soon in the COOK Report. value.

Introduction e-Science is a third leg.” Originally, through large-scale, complex as the Fourth there was just experimental instruments which relay data science, and then there was to data centers, and only then Paradigm theoretical science, with Kep- do they look at the informa- ler’s Laws, Newton’s Laws of tion on their computers.” Editor’s Note: Twenty years Motion, Maxwell’s equations, ago I was privileged to be and so on. Then, for many ‘The world of science has immersed for almost three years in Computational Sci- ence at the John von Neu- mann National Supercom- puter Center in Princeton. I understand the first three: empirical, theoretical and computational. The fourth is fall out from two more decades of Moore’s law – data intensive in that the computational simulations of twenty years ago are now extended in hitherto unimagined ways yielding experiments that produce huge amounts of complex data.

The following paragraphs are taken from a new book The Fourth Paradigm dedicated to problems, the theoretical changed, and there is no the memory of Jim Gray. models grew too complicated question about this. The new http://research.microsoft.co to solve analytically, and model is for the data to be m/fourthparadigm. people had to start simulat- captured by instruments or ing. These simulations have generated by simulations be- The introductory chapter of carried us through much of fore being processed by soft- the Fourth Paradigm is “Jim the last half of the last millen- ware and for the resulting in- Gray on e-science: A Trans- nium. At this point, these formation or knowledge to be formed Scientific Method” simulations are generating a stored in computers. Scien- based on the transcript of a whole lot of data, along with a tists only get to look at their talk given by Jim Gray to the huge increase in data from data fairly late in this pipeline. NRC-CSTB1 in Mountain the experimental sciences. The techniques and technolo- View, CA, on January 11, People now do not actually gies for such data-intensive 2007. Early on Gray says, look through telescopes. In- science are so different that it “Look, computational science stead, they are “looking” is worth distinguishing data-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 41 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 intensive science from com- the behavior of a cell or the what a temperature meas- putational science as a new, way a protein is built.’ urement is.’ fourth paradigm for scientific exploration.’ ‘This is similar to Jeannette Experimental Budgets Are Wing’s idea of “computational ¼ To ½ Software thinking,” in which computer science techniques and tech- ‘I have been hanging out with nologies are applied to differ- astronomers for about the ent disciplines.’ last 10 years, and I get to go to some of their base sta- ‘The goal for many scientists tions. One of the stunning is to codify their information things for me is that I look at so that they can exchange it their telescopes and it is just with other scientists. Why do incredible. It is basically 15 they need to codify their in- to 20 million dollars worth of formation? Because if I put capital equipment, with about some information in my com- 20 to 50 people operating the puter, the only way you are instrument. But then you get going to be able to under- to appreciate that there are stand that information is if literally thousands of people your program can understand writing code to deal with the the information. This means information generated by this that the information has to instrument and that millions Jim Gray 1944-2007 be represented in an algo- of lines of code are needed to http://en.wikipedia.org/wiki/Jim_Gr rithmic way. In order to do analyze all this information. ay_%28computer_scientist%29 this, you need a standard In fact, the software cost representation for what a dominates the capital expen- “We are seeing the evolution gene is or what a galaxy is or diture! This is true at the of two branches of every discipline, as shown in the next slide [Figure 2]. If you look at ecology, there is now both computational ecology, which is to do with simulating ecologies, and eco- informatics, which is to do with collecting and analyzing ecological information. Simi- larly, there is bioinformatics, which collects and analyzes information from many different experiments, and there is computational biology, which simulates how biological systems work and the metabolic pathways or

Sloan Digital Sky Survey was sailing is also well north (SDSS), and it is going to of the East-West ship channel continue to be true for larger- And Then Jim Gray used by freighters entering scale sky surveys, and in fact Vanished . . . and leaving San Francisco for many large-scale experi- Bay. The weather was clear ments. I am not sure that Despite human desires and that day and no ships re- this dominant software cost is especially those of scientists ported striking his boat, nor true for the particle physics to explain and know every- were any distress radio community and their Large thing -- just when you think transmissions reported. Hadron Collider (LHC) ma- you know, and are sitting chine, but it is certainly true astride the world with the On February 1, 2007, the for the LHC experiments.’ progress you are making, DigitalGlobe satellite did a fate sometimes rudely inter- scan of the area, generating ‘Even in the “small data” sci- rupts. So it happened with thousands of images. The ences, you see people col- Jim Gray. Suddenly and images were posted to Ama- lecting information and then tragically and mysteriously. zon Mechanical Turk in order having to put a lot more en- to distribute the work of ergy into the analysis of the We read “During a short solo searching through them, in information than they have sailing trip to the Farallon Is- hopes of spotting his boat. done in getting the informa- lands near San Francisco to tion in the first place. The scatter his mother's ashes, On February 16, 2007, the software is typically very his 40-foot yacht, Tenacious, Friends of Jim Gray Group idiosyncratic since there are was reported missing on suspended their search, but very few generic tools that Sunday, January 28, 2007. continue to follow any impor- the bench scientist has for The Coast Guard searched for tant leads. The family ended collecting and analyzing and four days using a C-130 its search May 31, 2007. The processing the data. This is plane, helicopters, and patrol massive high-tech effort did something that we computer boats but found no sign of not reveal any new clues.” scientists could help fix by the vessel. building generic tools for the http://en.wikipedia.org/wiki/J scientists.” Gray's boat was equipped im_Gray_%28computer_scie with an automatically deploy- ntist%29 “I have a list of items for able EPIRB (Emergency policymakers like CSTB. The Position-Indicating Radio Just when one would be cer- first one is basically to foster Beacon), which should have tain that the tools of Gray both building tools and sup- deployed and begun trans- and his colleagues would porting them ...” Fourth mitting the instant his vessel have found evidence. . . they Paradigm, pages xviii – xx sank. The area around the did not. Mystery seems to Farallon Islands where Gray have had the last word.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 43 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 V. Making E-science Work: The Middleware Solution An Interview with Bob Hertzberger

L. O. (Bob) Hertzberger is tor asked him to provide case in physics. The steam the Director of the Virtual some context around his new engine was developed based Laboratory e-Science project, approach to developing mid- on these insights. This, in better known as VL-e, an es- dleware at the VL-e. turn, laid the foundation for sential contributor to the Na- the industrial revolution and tional ICT Research Infra- Hertzberger: I would prefer railroads. Later the combus- structure for The Nether- to modify the Science Para- tion engine emerged that lands. digms slide in the previous helped to make the automo- section to observe that till bile and the airplane possible. Hertzberger leads the critical around 1930-1940 theoretical VL-e initiative to develop the science and empirical science …e-science must be functional model that, to- had developed to such an ex- gether with the methodology tent that complex theories able to combine of multi-disciplinary collabo- were laid out and complex data-centric and ration, is shaping a new gen- experiments were conducted eration of advanced e-science in order to understand the computational software. Hertzberger’s goal phenomena of empirical sci- is to provide generic middle- ence. Above all, this was the models. ware that is easily adaptable to many disciplines in order to make the technologies of “The Fourth Paradigm” affordable and accessible to many more researchers.

This is expected to provide a critical component in the Dutch effort to create a national knowledge infrastructure. But as can be seen in the interview with an Ameri- can medical researcher in Section IX, the new model and resulting benefits are having a global impact.

The COOK Report met with Bob Hertzberger at the Port- land Convention Center on November 17, 2009. The edi- Bob Hertzberger at the Portland Convention Center on November 17, 2009

By that time the interplay puter. The computer has like a microscope or a tele- between theory and experi- turned out the most flexible scope. That data could then ment necessary to under- instrument of all. be subjected to complex stand empirical phenomena processing, and the results had progressed in such a way In the beginning it was rather analyzed by humans to pro- that theorists developed clumsy and very limited. But duce findings that would give complex models. Experimen- as its speed and storage ca- more insight into, for exam- talists carried out complex pacity increased, the comple, the workings of a cell or experiments partly to verify puter became capable of the shape of a galaxy. Today, the theorist’s models, partly making far more complex the most powerful application to search for new phenom- calculations than a human of computer instrumentation ena. being. It could also store all occurs when we combine the kinds of digital information results from simulations with One of the best examples is more effectively. those of real experiments. the Bohr model for under- Doing so will yield better standing atoms and mole- The computer made it possi- simulation models. cules that constitute the vari- ble to experiment with mod- ous elements like iron, cop- els what we call simulations. The computer per, gold etc. (For lay per- Simulations could represent a sons: the Bohr model says certain phenomena — for has turned out to the each atom or molecule of instance the impact of CO2 an element is composed of a absorption in the atmos- be the most nucleus with electrons circling phere. It could be executed flexible scientific around it like planets around by the computer with a pa- the sun.) Before the Bohr rameter of interest as defined instrument of all. model various other models by the scientist (for example had been developed. But the impact of burning fossil Twenty five years ago we be- when these models were fuels that then release CO2 gan to connect computers tested with the combination into the atmosphere) set to a together into networks. of theoretical work and ex- certain value. After that we These connections resulted in perimental evidence, the re- could change the parameter, the emergence of the Inter- sults demonstrated they did run the model again and ob- net and Web. Over the past not behave as predicted by serve the effect of our decade copper wires compos- the empirical observations change. These types of mod- ing these connections are be- that most elements are sta- els have been designed for a ing replaced by optical fibers ble. large number of sciences in- making it possible to com- To carry out such experi- cluding physics, chemistry, municate by means of high ments, scientists have used and pharmacy. But modeling bandwidth at virtually the instruments such as the mi- and simulation could also be speed of light. This has re- croscope to study all kinds of used to make economics sulted in what we call small objects like the cell, or forecasts or study road traffic “globalization”, where data the telescope to study the or social behavior, etc. and information about eco- planets or stars or galaxies. nomics, society and business But in the mid-forties of the The computer could also be can be communicated and previous century there used in “real world” experi- shared globally on an un- emerged a new instrument ments to gather the data in a precedented scale. which was called the com- digital form from instruments

The Promise of “The are the basic foundation for ing a systematic approach to Fourth Paradigm” the connection of scientists to e-science as a whole, in order major instruments and sets to provide a framework for of data gathering tools. Such creation of the various soft- Data and information can networks must be global in ware tools necessary to real- now be seen as the core scope, and the national agen- ize an e-science environment. component of science, as the cies that support them must computer penetrates ever figure out how to cost effec- Via this model we intend both deeper into the profession of tively allocate funds for the to stimulate the creation of the scientist. Accessing, han- network, and the instruments communities of best practice dling, sharing and combining that the network ties to- and to enable multi- the resulting data and infor- gether, as well as the server disciplinary collaboration. We mation will become more and clusters that record the data also believe our model helps more important. Optical net- generated. to implement the hardware works make these processes and software environment easier and faster. …we develop a necessary to realize e- science. The consequence of the emergence of computers, es- systematic Our model delineates dual pecially when connected via approach forces. On the technology fast networks, is that a new side are the computer, stor- form of science is emerging to e-science age and networking environ- which is called “e-science”. It ment, including all software enables the practice of sci- And as Cees de Laat pointed needed to harness that envi- ence at a global scale by out earlier, the cost of the ronment like Web, Grid and means of communication software – which makes it Cloud. We call this “technol- over the internet. For this possible to do something ogy push”. form of science to be suc- useful with all the hardware cessful, it is necessary that -- can exceed the cost of the On the application side, we data and information pro- hardware. Software has too duced with public funds stay often been an afterthought. describe the instrumentation environment needed to real- public and that access (for The Dutch Roadmap for SURF ize experiments including instance via public data- gives the greatest emphasis simulation experiments. Our bases) be supported and fa- to the need for creation of application side includes all cilitated. It could well be that this e-science middleware the software tools needed to such an e-science computer stack on which you are work- acquire, process and analyze and network infrastructure ing. What is the story behind the data so that the informa- will turn out to be the most your new software? tion and knowledge neces- flexible instrument ever de- sary for the scientific under- signed. We are working in the Hertzberger: In the Virtual Netherlands to realize this Laboratory for e-science pro- standing is produced. We call this “application pull”. potential. ject we have tried to bridge The environment needed to the gap between a state of accomplish our work should COOK Report: Your context the art network and being not only contain methods and for Jim Grays’ last lecture is able to use such infrastruc- tools that we have produced also the foundation for your ture for doing new science. ourselves but it should also own work. Optical networks We have begun by develop- be capable of incorporating

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 46 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 tools made by other scien- sible of the tists. That is why, besides technology realizing the application pull push ex- model and researching new erted by methods and tools ourselves develop- we also have focused on ments in methods and tools needed to Web, Grid, incorporate existing e-science and Cloud tools. For instance we want computing. to be able to use the Taverna In practice, workflow tool [See 4th Para- however, digm p. 137] in our environ- t h i s i s ment. We have no intention of never wasting resources to re-invent completely the wheel. possible. As an example, when Grid The Maturation of Our Information and changed from a process into a service-oriented environment, Application Tools data become the we had to adapt to this serv- core business of ice orientation. Similarly as Hertzberger: The slide titled Cloud computing becomes VL-E Project on the third e-science… more mature and standard- page down represents the ized we will have to adapt our overall structure we are system to meet the demands building to become, as much While we concentrate on the as possible, independent of application pull of e-science, of this new technology push. These dynamics also hold for the Cloud or Grid. It is com- we remain dependent on the Web, especially as the ing into shape as a set of changes in the technology semantic Web matures and is tools represented by a lay- push. Ideally we want to be- incorporated into current Web ered structure. The purpose come as independent as pos- standardization. is to enable researchers to have the most powerful and effective software possible to connect their workstations to the resources, whether they are databases, scientific instruments or high performance computing located in the global grid or cloud.

To help researchers who use clouds and grids, the field has created numerous tools such as Carole Goble’s Tav- erna. Taverna is a very good example of an e-science tool. These tools are designed to be used by a large commu-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 47 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 nity to capture their methods be the study of system biol- their research in more open and process their data and ogy. ways? Wouldn’t that re- information. Now each com- search be more effective than munity tends to have its own COOK Report: While the a closed proprietary ap- sociology. software could be open proach? source, and most of the da- In life-sciences for instance tabases could be open, but Hertzberger: In the Nether- the emergence of new in- the processes and results lands some plant breeding strumentation has made so- have not necessarily to be companies are very afraid of called omics-based experi- open? Monsanto because they pat- ments (transcriptomics, pro- ent a lot of seeds. Monsanto teomics, etc.) possible. The Hertzberger: Yes. A phar- can do that because they omics technologies form the maceutical company may well have more powerful devel- basis of a new form of (mo- use open tools but as long as opment facilities of which im- lecular) biology that cannot all its critical knowledge and portant parts are bio- be realized without extensive data are wrapped into a pro- informatics tools. But I have use of computer technolo- prietary database under its told a lot of people not to gies. Its consequence is an own control, that company worry because if you build increase in complexity of the should feel secure. your own e-bioscience chain experimentation as well as you can compete with them. data and information that ar- Let me loop back to life sci- Why? If they keep their rives in greater quantity and ences and the pharmaceutical chains proprietary, they will complexity than ever before. industry with which I am have more software costs This has had a tremendous quite familiar. Here you and less flexibility to use impact on experimentation in would expect that their data- other people’s tools. We have biology, pharmacy and medi- bases about the clinical trials an example of a pharmaceu- cine. e-science as a data cen- which are their core business tical company in the Nether- tric science helps in harness- will be kept closed. However, lands who for decades, at ing the (computer generated) the entire infrastructure could considerable costs and not complexity impacting the ba- use open software tools. much success, has tried to sic methodology by means of Nevertheless, at the end, set up a proprietary e- which domain science (in this some of the databases that bioscience chain. It is now case life science) is accom- house their results, and the embraces the semi-open col- plished. management of their data laboration model I am advo- will need to be closed (pro- cating here. As a next step we have to tected) because they contain demonstrate that our e- company sensitive informa- An open environment can be science model can be used to tion. As long as companies far stronger, because many handle data from instruments can protect critical data, in- more scientists can contrib- as well as data generated by formation and knowledge, ute to the power of the chain, computational simulations. I there is no reason why public cross checking it and contrib- am confident that will be pos- and private research parties uting. Last but not least, be- sible. In e-science you need should not work together. ing part of an e-science envi- access to both data-centric ronment means that you and simulation models to COOK Report: What possi- don’t have to support all the help you understand scientific bility do you see that re- tools yourself, especially problems. An example would searchers will begin to do when they are public domain.

So I agree with you that sci- teach you more in a few days and shape their interaction ence will be a lot more pro- than you could learn from with the database telescopes, ductive when all these new e- books in months. I am work- other instruments and super- science efforts are brought to ing in e-science now because computers in cooperative ex- fruition. These ideas are I believe that you must de- ploration with colleagues similar to those of Tony Hey. velop tools that make this around the globe. He started the UK program in kind of learning, learning via e-science. Now Tony is at Mi- electronic means, possible for Hertzberger: That is cor- crosoft where he is working people in all fields of science. rect. In the functional model towards this end. Not just in high energy phys- we are using and that is pre- ics. Personal contacts are sented on next page (people COOK Report: Doesn’t Mi- essential in science, and tease me by saying that it crosoft have a proprietary those can be supported by e- looks like a multi-stacked attitude? There must be real science through tools for ocean-liner) the towers limits on how proprietary you supporting collaboration and standing on the horizontal can be and still be competi- structuring multidisciplinary slabs represent, in fact, the tive in e-science. interactions. different user communities. Collaboration with industrial The total technology chain Hertzberger: Precisely. partners can be realized via necessary to create the That is the whole point. public/private partnerships problem-solving environment Openness via e-science for pre-competitive research. a scientist needs for his work makes you, as a scientist, This can cover the company’s is made up of the software more competitive. But let me end domain core business represented by the towers offer another insight. That I (for instance, pharmaceutical, and all the underlying soft- have ended up in e-science at biological or medical re- ware and hardware layers. the end of my career is be- search). But it could also be This software could reside cause I had a splendid career in the form of R&D necessary both on the scientist’s laptop in high-energy physics. High- to enable an e-bioscience and be located elsewhere as energy physics could never problem-solving environ- well — for example in Web, even exist without interna- ment. A well-designed e- Grid or Cloud based servers. tional collaboration. This is science infrastructure will It should allow him to do his now coming up more and help everybody including experiment and connect with more in other fields of sci- those companies. fellow scientist when neces- ence as well. sary. But it requires in-depth Here we are talking about e-science research to find …openess via doing science on a global ways how to use it. level. Using an e-science en- e-science makes vironment you can build dif- Let’s assume we have a sci- you as a scientist ferent scientific communities entist working in the area of -- public, private or mixed. medical and cognitive appli- (or organization) cations, and furthermore that COOK Report: My attempt he wishes to collaborate with more competitive. at defining this platform a colleague in the U.S. work- would be to call it the set of ing in the same area. Now Have lunch in the cafeteria at infrastructure tools that be- imagine that they do an ex- CERN, and you will meet all comes the medium through periment in cognitive brain kinds of people who can which he or she can choose research collaboratively. They

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 49 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 use the green silo as the source or the “basis” of their software for doing the experiment. By means of the same software they can communicate with the database located somewhere else in which the data of some of their previous experiments as well as the results of that experiments are stored. To- gether, they use the tools necessary to have access to their shared data in the database and have the tools – perhaps workflow tools - necessary to complete a new study. After finishing that study, they can store their moves forward put increased very useful because it for- findings, which they could pressure on people to choose malizes the structure of the choose to make public or to open collaborative research? experiment — and conse- keep for themselves until quently make it more repro- they get a better understand- Hertzberger: Certainly, but ducible. ing of the issues. don’t forget that researchers in a big pharmaceutical com- COOK Report: And seman- If they decide to make the pany like Merck will also have tic capture becomes an inte- database public, they can an- access to the same infra- grated part of the tools? nounce to the other re- structure without the need to Correct? searchers in cognitive sci- publish things they want to ence: the results from our keep private. In short, while Hertzberger: Yes. From the study are here. Please have a the methods of doing re- beginning of the VL-e project look and comment. Or they search are changing drasti- we recognized that semantic could select others with dif- cally, the data and informa- modeling was of vital impor- ferent expertise and invite tion becomes the core busi- tance. Research involving the them to join their work. ness of the scientist in both semantic web has made public and private organiza- available knowledge repre- COOK Report: Then you get tions. This opens up new op- sentation mechanisms (for the widespread incremental portunities. But it demands in example OWL and RDF) that benefits that the Linux com- depth e-science to enable its have gained general accep- munity provides from every- application. tance, together with a host of one who is interested in the tools to manipulate semantic process and actively thinking These research methods can models. It is also clear that, about how to improve it. But benefit, for example, from given the speed with which while no one is going to force the use of a workflow tool to data are gathered in e- them to make their data pub- describe the experiment that science, in creating semantic lic, will the “evolutionary” must be done. Such a tool is meta-models, we must rely process that occurs as this

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 50 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 on adaptive techniques. For would certainly be good. collaboration. This is a big this purpose we created the obstacle for “open” collabora- AIDA toolbox: Hertzberger: Correct. But tion in those sciences that http://www.adaptivedisclosur doing that will be more diffi- still depend on a more tradi- e.org/aida cult with e-science, because tional publication reward sys- it makes science more open, tem. As an example: Semantic in- controllable and reproducible. formation is important in the System Level Science context of Workflow composi- COOK Report: Because it is tion, and this so even if reproducible, by definition it There is another consequence automatic workflow composi- is verifiable? of e-science which might tion remains a long term am- have an even more profound bition. In order to incorporate Hertzberger: But now we impact on science. This is the a service in a workflow, one could go a step further. It is fact that e-science supports a needs to know what the serv- the step advertised by Tony system level approach to- ice can do and what kind of Hey and is beginning to take wards science. In modern data it accepts and creates. shape in some communities. science and engineering, and This knowledge is essentially In those communities you in society, too, we are in- semantic information. electronically publish your creasingly faced with complex article together with the in- problems which can not be This work is similar to the formation needed to verify studied separately. They can example of Taverna by Carole for others that the experi- only be understood in the Goble. She has built a work- ment will have the outcome context of the full overall sys- flow toolbox that she hopes that you claim. A publisher tem they belong to. The the whole biological commu- will tell you, if you want to study of this type of problems nity will use and which will publish an article you also is often called as system level also capture semantic infor- have to publish the data and science. One of the best defi- mation. An other example is information on which that nitions for it is: “the integra- that semantic models will article is based. tion of diverse sources of play a large role in the sys- knowledge about the con- tems and network manage- In the future in biology pub- stituent parts of a complex ment in Grid and Cloud com- lishing an article almost cer- system with the goal of ob- puting itself. Automatic con- tainly will mean publishing taining an understanding of figuration of systems and the workflow and relevant the system's properties as a networks will be impossible data/information so that whole.”, from Ian Foster without them. other people in effect can (Argonne National Labora- “replay” the experiment. tory) in the November 2006 COOK Report: I like these However, we have not yet issue of scientific journal IEEE ideas very much. This past talked about the fact that e- Computer. summer I read about some- science demands for a multi- thing that would have been disciplinary way of doing sci- Some examples of system unthinkable in my youth. ence, which creates a ques- level science might be: Clinical studies in the New tion of who gets the reward. • study into the role of black England Journal of Medicine In the CERN example, new matter for understanding had been faked! It seems as rules have been put in place the origin of the universe though this would end those to give everybody his/her re- practices and doing that ward in a multidisciplinary

• study of the fundamentals of matter in physics • the role of understanding the functioning of the cell for system biology • cohort studies in medicine (biobanking-based) • environmental studies into the role of biodiversity in ecosystems • studies of water and air pol- lution • discovery of biomarkers or ligands for drug design in pharmacology • modeling the behavior of continental shelves for earth quake prediction a systematic approach to e- ence of developments in the science as a whole. We do project as well as the crea- Because of the potential of e- this to provide a framework tion of modular software science to gather information for creation of the various components. Our model must on a global scale, it becomes software tools that are nec- contain the problem-solving possible to utilize all that essary to realize an e-science environments necessary for data and/or information for environment. We have de- the different applications and doing system level research. veloped a functional model, application domains. In de- However, it then becomes shown on the “VL-e Func- veloping our model we made important to understand how tional Model” slide below, a distinction between those that information can be inte- with the intention of stimulat- components (services) that grated in order to add to the ing the creation of communi- can be realized as generic understanding of the complex ties of best practice as well and those components which problem under study. as multi-disciplinary collabo- are more application-specific. In other words, how do you ration. design system level models This makes it easier to modu- that are capable of integrat- larize software components. ing all that data and informa- …e-science is an It helps to create generic tion? enabler for system software where-ever possible and therefore encourage re- COOK Report: You talk level science, usability. And the goal – about the role of the contex- an example given the expense of soft- tual integrator. That seems ware development - is to absolutely critical. Wow is of which is make software more capable, your approach to e-science more powerful and more different from that of others? systems biology useable over a broader range of applications. Hertzberger: In the Virtual Additionally the functional Laboratory for e-Science (VL- model supports the coher- Focusing on modularization e) we have been developing

change in the generic software as well as the application specific parts. To build an application, you apply as many generic services as possible so that the application-specific part is as minimal as possible. Then if you have to change your application, you only need to change a small part. You can leave the rest untouched. Al- though all services try to access the underlying Grid or Cloud services through a well-defined interface, that does not guarantee that when major changes in the to lower cost also has the plication domains. All soft- underlying software take effect of influencing our other ware has to be realized via place, such changes do not applications to become more the Software As a Service have an impact on your total generic. Workflow would be (SAAS) concept. It’s impor- system. an example. tant that all application services be designed as much as COOK Report: Abstraction We have already explained possible to be independent layers are determined by that in realizing such a func- from changes taking place in communities of best practice. tional model, one has to ob- the “technology push” envi- As more of these communi- serve two important devel- ronment. As an example, ties emerge, will this require opments: the push of the suppose the cloud is trans- adapting the functional new technologies, and the formed into something else. model? changes taking place in those Then we have to modify technologies. On the other those modules that are de- Hertzberger. Yes, we must side you have the pull – in signed for communication then refine the model. Proba- other words, the demands with the cloud environment. bly we will add more abstrac- from the various application tion layers. It took a lot of domains like physics, medi- COOK Report: So the ge- trial and error before we saw cine or pharmacy as they are neric modules sit on top of we had the current layers placed into your total e- the evolving foundation? Can correct. For me the proof of science system. they maintain their viability the pudding is always in the On the “application pull” side independent of what goes on eating. I needed to build the we minimally distinguish two above or below them? software to see whether the different abstraction levels. model was correct. One contains the application Hertzberger: The change in specific software. The other Grid from process to service The model we discussed is contains services that are orientation we talked about presented in a general way generic for all e-science ap- earlier required a major with a clear distinction between generic and application

studies, as developed within VL-e, is presented. Although it was used by the bioinformatics group within VL-e for prokaryote genomic studies, parts can also be used for eukaryotic transcriptomics research.

On the left hand side of both diagrams the general description of the type of software is illustrated. While on the right hand side the diagram shows some examples of tools that are present in that particular abstraction layer.

On the first diagram we can specific layers. I wants to tomics research, i.e. genome- observe that, in order to en- use it in this discussion to wide expression profiling ex- able these studies, a new ab- illustrate that because of the periments, he needs to have straction layer is introduced work we’ve done, we already a transcriptomics Problem being a domain generic e- believe that more generic Solving Environment with the Bioscience layer. This is a layers are possible. Consider correct tools. A workflow tool layer that’s generic for a spe- the work going on in VL-e will be part of that. In the cific life-science e-science currently to support one of two diagrams, the Genius en- research domain. In this par- the basic genomics technolo- vironment for transcriptomics gies. These are studies based on micro-array instrumentation. They’re called transcrip- tomic studies. This is a specific instance of the general process of how we build problem solving environments in VL-e. The process is similar whether it’s for biodiversity or food or medical or cognitive research.

A Technology Chain for Micro-array Tanscriptomics

Hertzberger: In order for a scientist to do transcrip-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 54 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 ticular case, the necessary kinds of tools like the imag- have brought into our ap- generic tools for prokaryotic ing tools mentioned above. proach. If you would like to transcriptomics studies are In both the illustrations Tav- build something for pro- already present erna is a workflow environ- teomics or for metabolomics, ment. “R” is a workflow tool you could use already a con- http://www.microarray.nl/pro that you plug into Taverna. siderable number of the tools jects_inf.html. CDF merger is a generic from the illustration and add workflow, that can be used in domain specific and addi- But it is easy to imagine that many environments for tran- tional domain generic tools as when an imaging detector scriptomics. The CDF merger they are needed. like a functional Magnetic is also one of the tools that is Resonance Imaging (MRI) used to get the correct and For each (sub)domain in the detector would be a require- up-to-date annotation of ana- life sciences, you want to ment for a study, imaging lyzed genes and related have a community that will tools should also become part pathways involved in biologi- reuse tools whenever possi- of this domain generic layer. cal processes. ble. These are present in the domain generic layer, called In the second diagram (to http://www.systemsbiology.n e-Bioscience in which the e- the right) the VL-e functional l/datgen/transcriptomics/tran Bioscience tools reside. model is presented again but scriptomics.html These tools can be used to now with the one extra e- develop Problem Solving En- Bioscience layer added to the Finally we have VLAM which vironments (PSEs) for differ- stack of abstraction layers. is a workflow environment we ent e-Bioscience subdomains developed before starting Examples include Biobanking We have chosen workflow current VL-e project. (the environment necessary tools as an example for the to do cohort studies in medi- diagrams. However, the lay- The illustration shows the ad- cine and biology), bio- ers also will contain other vantage of the modularity we diversity (the study of the developments and presence of different species), e-Green Genetics (the study of plant breeding), and e-Food (the study of for instance why is something tasting bitter or the study of food safety) All these application subdomains are presented in the illustration. We believe that in future you will discover that there are many more generic functions that can be reused than you expect.

COOK Report: What you seem to be saying then is that the domain specific layer in the GENIUS slide above

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 55 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 may be repeated in the dia- more a tool is applied for dif- automation (like office or ad- gram below to branch off fur- ferent research to solve ministrative automation) that ther upstream into more de- slightly different problems, the automating a domain process tailed research tributaries us- more generic and powerful it always requires a better un- ing slightly different tools and becomes. This is why I view derstanding and consequently plug ins into these other fours myself as being halfway be- a rationalization of that proc- areas below? tween being a computer scien- ess. This is a necessary first tist and an experimental scien- step towards further stan- Hertzberger: Yes, and what I tist. A computer scientist dardization of the domain am also trying to say is that wants to build as many reus- process in order to make it within such an area you will able tools as possible. An ex- suitable for automation. find new subdomains that are perimentalist says I have to also reasonably generic. The solve my problem even if it Participating in e-science, re- specific parts will be pushed means risking reinventing the search will force domain higher and higher up, and the wheel. And by the way, when communities into better reuse of software will be you make the tools modular, thinking through their re- pushed further and further they become easier to re-use, search problems and conse- along as well. But this is not because they can be applied to quently better understanding something that I can prove at solve different problems. their core business. This is a present. first step towards rationaliza- The more modular tion and will help in further COOK Report: But you ex- standardization of their re- pect to discover that as you you make your search methodologies. This find generic components in software, on its turn is a prerequisite your new research they be- for exchanging data/ come reusable in more and the more successful information an essential con- more operations. dition for collaborative re- you will be. search. Hertzberger: Yes, and the reason why I think I am right There is also a philosophical User Experience with is that the more one comes to point here that seems essen- e-Science understand about new e- tial. If you are going to use a The biologist responsible for science based experiments, computer in your experimen- the development of the tech- the more we find additional tation, (and this will happen nology chain for microarray generic aspects to those ex- more and more,) you must transcriptomics likes to say we periments than realized be- ask whether the methodology are now going into the fourth fore. And of course, in such (design) of the experimenta- phase of e-science develop- cases I also have the option of tion is not determined by the ment in the Netherlands. The using software components fact that you use that com- first phase was when most ap- already designed by others. puter. This implies that e- plication domains were just science is just another form of science and not only an en- learning all the stuff we’ve In addition functionality can be mentioned above, and realiz- abler making domain science added to make them better ing slowly that they had to possible. suitable to the problem at change their way of doing sci- hand. In other words, we can ence. Regrettably, it has to be Apart from this, one can also make them more generic. The said that for the most, only learn from other fields of

respect to scientific competitive advantages are so apparent. In this case inclusion of e- science in the basic curricula of biology and medicine will also help.

The VL-e Experimentation and Software Production Infrastructure

Hertzberger: As I mentioned earlier, for me the proof of the pudding is in the eating. Therefore we constructed two experimentation those scientists that were ac- applications this means that environments that acted as tively involved in the e-science we have to produce functional the software production and developments currently under- PSEs in order to prove the e- test infrastructure of VL-e. stand the challenges we are science concept. On the right hand side of the facing. For most life-science slide on the next page is the scientists, the computer is still …experimentation rapid prototyping environ- just an interface to rather ment for software develop- simple software tools. is the only way to ment, while you will see on In the second phase, we verify our ideas… the left hand side the proof of learned about the layered concept environment desig- structure and the fact that we Now we are indeed entering nated to test out and get ex- needed to build Problem Solv- the fourth phase. Here, we perience with application ing Environments (PSEs) with have to not only build these software. In addition the generic elements. In this PSEs, but also to support them diagram shows a test and phase the e-science approach in a structured way in order to certification environment that became immediately ex- make them successful. These helps in moving software tremely complicated and frus- PSEs are so complex that they from right hand experimental trating. need expert support. Other- environment to the left hand wise life-science domain scien- proof of concept environ- In the third phase, we better tists actually doing the ex- ment. understood that there was periments cannot use them. something as a domain- So for all involved application COOK Report: Therefore generic layer. And more im- scientists, this e-science quest the VL-E Infrastructure dia- portantly, we all realized our has been a difficult experience, gram that follows is very im- roles in the whole VL-e com- with little immediate reward in portant. It describes the plex. We should work from the their application domain. How- process that you use to ma- perspective of our own exper- ever, no one even considers ture your software from an tise, and our objectives and abandoning this approach, as experimental environment results should reflect that. For the long-term rewards with into a certification one, and

then into what you call proof ings to detail how your proc- them down. of concept, before being re- ess works. leased to first the community Hertzberger: I see. We call of different user in the VL-e Hertzberger: On the “VL-e it rapid prototyping, and this project in the form of the Infrastructure” slide at the rapid prototyping is exclu- proof of concept. Moreover extreme right in the red box sively used by computer sci- the “e-Science Roll Out” slide you see the words “VL-e ex- entists. The current system 2 pages down shows how the perimental environment.” (DAS-3) is a distributed sys- software is further made This is our rapid prototyping tem with multiple 10 Gb/s available to user communities environment in which scien- links (lambdas) between the spread all over the Nether- tists can develop their tools clusters, provided by SURF- lands via the Grid based ICT without having to take into net. research Netherlands infra- account that the rest of VL-e structure called BiG Grid. exists. They can play. It provides an excellent experimentation environment Hertzberger: Yes. And the COOK Report: This is what for computer scientists. To- number of BiG Grid users is Google and other folk would gether with the application growing steadily. call a “Sandbox” on the sense people, they can decide on a of the contained box of sand tool that they want to en- COOK Report: Take me then in which children play and hance, or they can build a on a tour through the draw- build things and then tear new tool.

Let’s assume that fore they can do this they the computer first must go through the test scientists and and certification environment application scien- that is shown in green above tists want to de- on the left side. The reason velop a tool to is that the tool now has to use in medical become an operational part applications. of a larger software environ- They take the ment. tools available to them in the right COOK Report: The certifica- hand red column, tion process is to see how it and they use fits with all the other pieces them to develop of the environment. their new tool. Hertzberger: Correct. It By the way, the three green Then at a certain moment must fit into and interoperate and turquoise cylinders that they are happy. They say, with the virtual laboratory you see sitting on the Virtual “OK this tool can now be tools and the grid middle- Laboratory on the left side, used in our medical applica- ware. When it passes certifi- conceptually should also be tion user environment.” cation it comes into proof of shown sitting on the red col- concept on the left hand side, umn at the right as well as on COOK Report: In other which completes the migra- the green column in the cen- words they release it to tion process. The scientists ter. But that would expand medical scientist end users to are using it, and based on the size of the diagram too play with? their experience, we decide much. whether it needs modifica- Hertzberger: Yes, but be- tion.

We developed something is part of the proof of the by the fact that people from called the V-browser that pudding and the proof of the government agencies moni- went through this process a pudding is – as I have said – toring our project in the be- number of times before the in the eating. If nothing else ginning complained that this users found it acceptable. this should have been what activity was reducing the This V-browser is a tool de- we promised to deliver to our amount of manpower that signed to make the study of government. That there was could be devoted to research. medical information easier some difficulty in understand- Now let’s look at the words and is now being used in ing what we did is illustrated “e-science roll-out” in the left medical communities. An- hand column of the slide that other example is the Java- …transferring our shows the entire process. GAT, which hides the different underlying grid middle- proof-of-concept COOK Report: The eating of ware. JavaGAT is also used software went the pudding is found in the by other international pro- turquoise colored proof of jects (for example TeraGrid well, most of it concept above? and D-Grid) and has also led to the SAGA standardization ran without Hertzberger: Yes. And effort. modification… when the eating of the pudding was declared good by For VL-e the proof of concept the VL-e user community, it

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 60 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 was transferred to BiG Grid separate organization with its want it to do? where it is used by about 40 own funding that will run for user communities now. BiG another two to three years. Hertzberger: Correct. And Grid is a separate project or- It will support other applica- they may even use a differ- ganization based on a pro- tions besides those of VL-e. ent model. Let me conclude posal that was submitted by And those folks will make with the description of the Nikhef (the Dutch institute new applications that run on VL-e experimentation envi- for subatomic physics), the top of ours. In transferring ronments. The proof of con- Netherlands Bio-Informatics our proof of concept software cept now resides with BiG Center (NBIC) and the Na- to BiG Grid, it turned out that Grid while the rapid prototyp- tional Computing Facilities most of it ran without the ing environment still runs on foundation (NCF). In 2006 need for any modification. the Distributed ASCI Super- the project was awarded 29 computer (DAS) of the com- million euro by NWO. BiG COOK Report: So what you puter science research school Grid was established in order are saying is that the middle ASCI. Big Grid can use the to set up a grid based user bars, Grid middleware, and Test and Certification envi- infrastructure for our scien- virtual laboratory transferred ronment in the future for its tists. The Dutch National from the VL-e proof of con- own software. Computer Center SARA was cept into the e-science roll brought on board as an op- out within BiG Grid. So you …funding is not erational partner providing handed things over from VL-e grid operations and support to BiG Grid as direct work on unlimited… to the BiG Grid communities. VL-e stopped? One major part of the BiG decision-makers Grid activities is the running Hertzberger: Yes, the proof had better invest in of the Dutch Tier I facility for of concept software will be LHC, which is operated jointly further developed, but at this infrastructure by Nikhef and SARA. point it will be done within first. BiG Grid.

Generic software This kind of development COOK Report: And in 2010 mechanism was always an should be the e-science Research Cen- idea I had and strived for. I ter that replaces VL-e will be did not expect the migration understood to be free to do whatever people an essential part of basic infrastructure

COOK Report: In other words BiG Grid is the railroad tracks on which VL-e’s e- science middleware runs?

Hertzberger: Exactly. You have seen that BiG Grid is a

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 61 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 to run so smoothly, and of the base software suites. Because if a user discovers those responsible for it did a As a scientist you need to something sufficiently ge- wonderful job. But on the have tool flexibility. And the neric, the engineer support- other hand, we did have a problem was that changes to ing him can integrate it into problem because a lot of the the tool boxes also impacted the toolbox he is using. tools in some of the tool the ability of the toolbox to These are things that we did boxes are not part of the interface with the virtual not try to design in advance proof of concept environ- laboratory services. but learned as the project ment. The domain generic went forward. In practice we tools are represented by tur- The key question for us was have either generic tools and quoise color bottoms of the how do we support the tools? especially domain generic cylinders below. But note The answer was that support tools, or we have application that as the color shows, they came primarily from people specific tools. And these run overlap and extend into the in BiG Grid. What I am try- in the BiG Grid environment. virtual laboratory as well. ing to explain is that the Depending on which commu- structure we have built is nity you are talking about The reason is that the valida- more flexible than it looks to they will use certain tool- tion process was designed to be in these figures, which boxes and certain tools. The make software supportable in give you just the impression more modular you make your an operational ICT environ- of what is generic and what is software the more successful ment. Whereas that approach application specific. In actual you will be. is essential for the base infra- practice what happens is that structure software, it is less someone builds his own tools The Economic well suited to software and and then asks for support Impact toolboxes supporting the core when those are used by a scientific process. Here, sci- community. The Committee of Wise Men entists were always making in the Netherlands was re- changes to their tool boxes, COOK Report: It sounds sponsible for giving us 20 because they got new ideas. like the boundary between million euros and we had to When they do they must virtual laboratory and domain find another plus 20 million adapt their tool boxes. generic tools is fuzzy? euros matching. We had to define targets. The most im- What we are doing now is to Hertzberger: It is indeed portant was to show that we cope with this problem by, on fuzzy. This fuzziness first had impact at the end. the one hand, making tools seemed to be a failure, but is as generic as we can. But on in fact a big success. Some- What is illustrated with this the other hand, since many times things are running in last slide is that in some of are 'domain generic', a more the virtual laboratory without the cases we built applica- agile deployment of toolboxes having obtained certification tions and tools, which if you based on community- from the BiG Grid organiza- had to build them from supported software distribution supporting it as an op- scratch, would already have tion has proven more effec- erational service. This is pos- cost you more than our total tive. In BiG Grid a team of e- sible because the user com- project budget. I will give you science support and devel- munity gives the necessary an example. In VL-e we built opment experts helps re- support. Therefore the users a virtual lab for ecological searchers to achieve this ag- are happy. Then you have, as analysis based on our generic ile environment built on top a result, a lot more flexibility.

the advantage that it was not necessary to start from scratch. We had domain as well as e-science knowledge available. Moreover, we pos- sessed the necessary e- science (proof of concept) environment on which to build.

COOK Report: Are the figures on the right side – 20, 10 and 15 adding up to 45 the projected costs of doing those tool sets from scratch?

Hertzberger: Yes and some people claim that I was far too conservative and that the tools and some specific ones. Euros. This is presented as actual from start costs would This is now in use by munici- the spill over explicitly called be two or three times higher palities, project developers out in the Economic impact than my figures. Somebody and building companies in the slide because the roots are all retired from IBM estimated Netherlands. Before they can in VL-e. the cost would be more than start construction activities, three times my figures. But I they have to make an inven- Another tool that we built and am a scientist so I chose con- tory of rare plants and ani- is now in actual operation is a servatively. mals living on the land where bird avoidance system for our they wish to build. They can air force. By combining a COOK Report: Because you use shared data and method- sensor network and models, have leverage – a multiplier ologies or tools to find out it predicts where flocks of effect? what species there are and birds are going and advises what they are allowed to do pilots how to avoid them. The Hertzberger: I’m happy to or not to do with each. Cur- same system is also used for say, yes, we have a multiplier rently about 50 people are scientific research into migra- effect. working on this project with a tion patterns of birds. Devel- total budget of 20 million oping these tools in VL-e had

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 63 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 VI. Growing E-Science Domains for The Netherlands Roadmap for a Next Generation of Science

“Dutch e-Science efforts tify those common features Editor’s note: – Biodiversity are at the leading edge” and deploy one infrastructure was just the first topic listed that takes care of these here. The remaining are Bio- While the development of e- common problems banking, Metabolomics Bio- science and related infra- molecular informatics, Cogni- structure is a global effort, [Editor’s note: the e-science tive science and have been Dutch scientists have been at interview Chapter V above moved to the text box below. the leading edge of these de- makes this point. Note also velopments. The VL-e mid- that the following abbreviated Physics, astronomy, term review Committee con- catalogue of Dutch e-science chemistry cluded in its report: effort is a shortened summary of an extensive de- Molecular simulation partment store like inventory. “The Committee considers Molecular Simulation has the Here are top level highlights. VL-e to be an excellent pro- potential to play a central A text box follows with subdi- ject at its mid-life point with role in the design of new ma- vided areas in life sciences high quality contributions. It terials and processes, and in has pushed back the frontiers and physical sciences.] the modeling of biological of e-science, whose aim is to processes. [snip] Editor’s induce a paradigm shift in the Life sciences note – Molecular simulation methodology of science itself, was just the first topic listed a difficult and worthy chal- Biodiversity here. The remaining are Par- lenge.” A variety of organizations col- ticle physics, Astronomy Ca- lect samples of species and talysis for energy conservation, As demonstrated by the make observations of their Fluid dynamics and have been cases in this section, the in- occurrence and abundance. moved to the text box below. frastructure elements needed These collections serve not by a wide variety of sciences only the study of agriculture, Climate research are very similar. All science water quality, bio- prospect- cases require a rich land- ing, conservation, but also Climate induced weather scape of resources, consisting research in systematics, bio- extremes of networks, computing, geography and population The consequences of global visualization and storage, but ecology. These data are more warming for Western Europe foremost services that are and more appearing elec- include an increased prob- instrumental in integrating tronically, scattered around ability for hot summers and this infrastructure and that various labs and institutes, higher winter precipitation, enable the scientist to effec- whilst also massive amounts the latter leading to an in- tively address and use it for of spatially explicit data on creased amount of water to his research. The most effi- the collection sites have be- be transported through Dutch cient solution to this e- come available. [. . . .] rivers to the North Sea, chal- science challenge is to iden- lenging the Netherlands for

These photos from The KAUST Exhibit at Supercomputing 09 indicate the use of 3d visualization tools where two scientists can explore protein folding. In this instance they walk toward the display with a visor and a pointer that enables them to steer the display. They can communicate in real time examining the same model from different perspectives. The second picture shows a virtual navigation menu by means of which the user on the show ﬂoor can change his interaction with the display which uses eleven “tiled” LCD monitors, nine of which are visible.

decades to come. It requires however, it is necessary to go For better accuracy, we even a combined investigation of to at least 40 km which will need to go beyond that. the underlying atmospheric require a more than 100-fold and climate processes. Simu- (10*10) increase in required Arts and lations with an atmosphere computing time. The same Humanities model with a horizontal reso- factor 10 holds with respect Services for the e-humanities lution of 400 km are possible, to ocean model resolution.

Life Science Detail Physics, Astronomy, Chemistry

Biobanking Detail The development of a (federated) biobanking database infrastructure as a critical component for biomedical life sci- Particle physics ences research and discovery is urgently needed. The ex- Particle physics has made striking advances in describing plosion of genotypic and phenotypic data requires that data the intimate structure of matter and the forces that deter- are properly stored, accessed, managed, queried, analyzed, mine the architecture of the universe. Nevertheless, funda- and shared with others. Flexible data integration is needed, mental questions like: ‘What is the origin of mass?’ and both across locations, and across application platforms, ‘What happened to anti-matter since the Big Bang?’ remain. varying from genetic, clinical, molecular, and demographical Parts of the answer to these and other questions are likely data, to analytical and biostatistics data [snip] to come from the Large Hadron Collider (LHC) at CERN. [snip] Metabolomics Metabolite profiles reflect biochemical changes, which Astronomy should be properly identified and analyzed such that the LOFAR is an innovative effort to force a breakthrough in different type of variations, i.e. the changes due to disease, sensitivity for astronomical observations at low radio fre- nutrition, life style, etc, can be differentiated. However, the quencies. It will explore the low frequency (10-200 MHz) field of metabolomics is challenged with the huge amounts radio sky at high resolution for the very first time, focusing of high-resolution mass spectrometric data produced and on several scientific goals: the study of the very early uni- the lack of an advanced bioinformatics framework for proper verse, the study of exotic phenomena such as the formation handling and pre-processing these data. [snip] of massive black holes, and (clusters of) galaxies, observa- tion of Gamma Ray Bursts, radio supernovae, intermediate Biomolecular informatics black holes, flare stars and exo-planets, and the detection The Center for Molecular and Biomolecular Informatics of extremely high-energy particles. (CMBI) has long been involved in projects aimed at improving the precision and accuracy of three-dimensional coordi- Catalysis for energy conservation nates of macromolecular structures. This is an important On a global scale, energy conservation and ecology-friendly aspect of a larger project aimed at the elucidation of the industrial production processes have captured much atten- molecular origin of genetic disorders. Late 2007 a major tion. Examples are the development of effective detergents step was set when all protein structure files that were ever at low temperatures and the development of low-energy solved worldwide were re-refined using modern software processes in the oil industry. An important area in chemical and today's understanding of geometric and energetic as- science copes with these phenomena and problems. [snip] pects of protein structures. This was a major operation that involved hundreds of computers spread over Europe but Fluid dynamics combined in the EMBRACE Virtual Organization for biosci- Fluid dynamics is the scientific discipline that covers all ences. kinds of flow phenomena: hydrodynamics, aerodynamics, climate research, meteorology, combustion processes, etc. Cognitive science With the presence of turbulence, flow processes in general It is widely believed that the structure of the human visual are extremely complex. [snip] systems is related to (statistical) regularities of the visual world. Understanding and simulating the human visual systems therefore requires understanding these regularities. . To this end, cognitive scientists from the University of Am- sterdam study properties of the visual world and their influence on the structure of the human visual system using whole-brain modalities such as M/EEG and fMRI [snip]

Digital humanities will profit content analysis (e.g. (semi)- isters), economic time series, from infrastructural support automated transcription of and geospatial data require for the handling of multime- spoken word collections, se- new approaches in fields such dia content in the interpreta- mantic clustering, semantic as phonetics, sociology, tive processes that are the video annotation), that econometrics, demography, heart of humanities research. stimulates community build- history, geography and ar- In order to allow individual ing and that can be coupled chaeology. DANS, the na- scholars and groups to share to visualization tools.[snip] tional center for permanent source data and the annota- access to research data for tions that are the result of Large digital collections of the humanities and social interpretation or comparison, annotated speech and other sciences, is exploring the po- the humanities need ad- language data (text, audio tential of the data grid and vanced collaboratories: in- and video), data bases on the linkage of heterogeneous teroperable platforms that population characteristics data resources.” can support and integrate (census data, population reg-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 67 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 VII. Potential Customers with Global Agendas An interview with David Zakim, MD

Dr. David Zakim wants But whereas there is already of health care and reinvigorating globally available multi- too much to learn, there is clinical research. disciplinary databases too often insufficient knowl- to help improve individ- edge for accurate prediction http://www.idm-foundation.org ual patient care in real- of risks and outcomes. Thus, time. everyday medical practice Medicine is struggles with two significant David Zakim is the kind of knowledge problems: too practiced on the “customer” that will increas- much to learn but not enough false premise that ingly need the tools that Bob to practice efficiently. Hertzberger’s team is devel- What’s missing, Zakim real- the physician can oping in the VL-e and the re- ized, is technology that in- search infrastructure de- hold all the sures the application of rele- scribed earlier in this report. vant evidence to manage- knowledge he will Zakim has a big agenda. He ment of clinical problems on wants to transform the prac- a day-to-day basis and a ever need tice of clinical medicine, clinical research infrastruc- which cannot be accom- in memory ture to support better under- plished without leveraging e- standing of risk factors for science. disease and the therapeutic In 2000 David Zakim retired efficacy of a variety of treat- from Cornell Medical School ments. after a 40 year career in aca- These are not new problems. demic medicine. But he was 30 years ago I published an unable to stop thinking about article in the June 1979 Fu- the chronic problems that turist about the early work of limited the ability of physi- Dr. Lawrence Weed on apply- cians to deliver effective care ing computers to these medi- to their patients. cal issues. Not much has Zakim knew that physicians changed in practice of medi- rely on memory to treat pa- cine, however, including that tients. The physician pre- it consumes an ever- tends he or she can carry increasing percentage of around in memory all the GDP. But technology has David Zakim MD medical knowledge needed to leapt ahead. match presenting problems Zakim’s CLEOS system is de- Zakim started work on these and formulate a defensible signed to change dramatically issues upon retirement from diagnosis and treatment plan. the acquisition of patient his- Cornell and founded the Insti- tories, the assembly of pa- This is a physical impossibility tute for Digital Medicine in 2007 tient medical records, the co- because of the enormous as the organization to fulfill his ordination of all personal scope and complexity of the vision for improving the quality knowledge base for practice. medical data gathered over a

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 68 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 patient’s lifetime and the been implementing a com- am trying to achieve in a assembly of very large clini- puterized questionnaire and specific knowledge domain.” decision-tree-based system cal databases that describe [Editor’s note: At the start for collecting and interpreting the behavior of disease of his presentation Neggers clinical data via direct inter- across time. describes the SURF Roadmap action with patients in the Towards a National ICT Re- absence of physician input. search Infrastructure devel- …Zakim realized That work is described in the following article in the litera- oped jointly by the ICT Infra- what was missing: ture. structure providers in The real-time Netherlands described earlier http://www.biomedcentral.co in this report]. databases that m/1472-6947/8/50 Most of this book has focused instantly and The real-time capabilities for on the practitioners, proc- clinicians envisioned by David esses and “products” of ad- continually Zakim, however, will require vanced ICT infrastructure re- a global ICT research infra- improve patient search in the Netherlands. structure. It will need soft- Before going to press, it management ware tools for multi- seemed a good idea to spend disciplinary teams, capacity some time talking to a poten- Zakim’s goals are three-fold: for massive databases, fed by tial user as well. enable a computer system to millions of heterogeneous collect, analyze and make sources, well-oiled database Here are highlights of a Janu- available the broadest possi- integration, with robust and ary 27 follow up to an earlier ble range of patient data, agile networks. All this will two hour interview with Dr. from patients around the necessitate massive real-time David Zakim. world via direct interactions processing and data transport with people seeking health capabilities. care; assemble large clinical Zakim needs specifically the COOK Report: David, what databases that are amenable kind of e-science tools, teams is your reaction to Kees Neg- to mathematical analysis; and work processes that Kees gers presentation on what and allow delivery of man- Neggers, Bob Hertzberger SURFnet and Bob Hertzberger agement advice through real- and their many colleagues are doing. time comparisons of data for have spent years evolving. individual patients with the Recently Zakim was emailed Who is Neggers and accumulated clinical experience within a world-wide an excerpt from the presen- how fast can I meet clinical database. tation Kees Neggers gave at NORDUnet in September, him? In cooperation with Dr. Mark 2009, plus a brief overview of Dominik Alscher, Professor of the goals, strategies and Zakim: The first of Kees Internal Medicine, the Uni- achievements of ICT research Neggers’ recent research versity of Tuebingen, Ger- in the Netherlands. trends is system level sci- many and director, depart- “Who is Neggers?” Zakim ence, which is defined as the ment of Internal Medicine emailed back. “And how fast and Nephrology, Robert integration of diverse sources can I get to meet him? His Bosch Krankenhaus (RBK), of knowledge about the con- Germany, Dr. Zakim has paper describes the problems stituent parts of a complex and aims of exactly what I system. This describes some

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 69 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 areas of science, but it also made is the importance of come of interest. applies to the entirety of multi-disciplinary research. medicine. Our problem is a Physicians are already in We know smoking contrib- pediatrician knows very little some form doing multi- utes to coronary disease. So about adult medicine; interdisciplinary research already. we might decide that algo- nists essentially know zero Epidemiologists, who don’t rithms for predicting the inci- pediatrics. An obstetrician treat patients, use MDs ex- dence of coronary events can’t take care of a patient amine large databases to dis- need to be improved. Conse- with heart disease, nor can a cern cause and effect. quently we’re going to sub- cardiologist take care of a define smoking as one way of pregnant woman. Even within We need a far more robust determining a better algo- the specialties of adult medi- form of multi-disciplinary re- rithm. Now that excludes all cine, such as internal medi- search. In clinical research other data and is based on cine, a gastroenterologist will we need non-medical exper- the idea of what we know, know little about the diagno- tise such as computational i.e. we know smoking is im- sis and management of car- expertise and mathematical portant. But it could be that diac disease. expertise. Within the disci- eye color is important as pline of medicine, examina- well. No one has ever COOK Report: The human tion of clinical data is based checked. being is a complex system. exclusively on concepts of patho-physiology. (Neggers It’s impossible to look at all Zakim: Enormously complex. refers to exploring data sets possibilities, except using Because of that complexity, with upfront hypotheses.) But patho-physiology, unless you and because of the scope of we know the power of that posit the question that we knowledge available, medi- has reached its limits. have to begin to look with cine has been increasingly abstract mathematical algo- cut into smaller and smaller COOK Report: Doesn’t an rithms. In that case we need specialties and subspecialties. upfront hypotheses bias your to involve mathematicians in Specialists do not communi- thinking and choices? examining very large data- cate with each other; they bases. They can examine da- just refer patients. …powerful tabases of unlimited size and do it with powerful algorithms Medicine, in Neggers’ terms, algorithms can that don’t depend on pre- is system level science. To formed ideas of cause and capture the knowledge base, analyze databases effect. Almost all epidemiol- to enable a machine to help of unlimited size. ogy is based on the idea that an individual to practice X causes Y, and then prove it medicine, one needs to inte- or not prove it. grate knowledge from diverse Zakim: More than bias, it groups of physicians with dif- says we’re going to look for a Let’s say we have a factor ferent subspecialties, differ- relationship between this set with 10 possibilities. And ent knowledge sets, and a of data and some outcome. none of them appear to be wide range clinical experi- The dataset is cherry-picked important. None of the com- ence. according to pre-conceived binations turn out to be sig- notions of what kinds of nificant. But that doesn’t The second point Neggers events might impact the out- mean that one of them is not

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 70 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 important. It’s simply that port that sort of approach. ownership of programs like you couldn’t show their im- CLEOS would transfer devel- portance in the context of COOK Report: Why does the opment and deployment of how you examined the data. Netherlands’ system appear new medical knowledge from They might, however, turn so promising? the medical profession to out to be important in the profit-driven entrepreneurs, context of 100 or 1,000 data Zakim: In Netherland they who would be guided by fields when relating to the have set out to develop soft- ‘share-holder value’ not the outcome, let’s say, of a coro- ware that can support that public interest.” nary event. sort of approach. No one else Zakim: The third thing Neg- that I know has that. I read gers emphasizes is real-time The CLEOS system developed about a company in North processing of very large da- at our Institute for Digital Carolina that builds most of tasets. At the current time medicine allows us to popu- inventory databases for big there is no real-time process- late “yes or no” in approxi- corporations. They’re innova- ing of medical databases. mately 18,000 to 19,000 tive but of course their ap- Typically if a databases ex- data fields. Mathematicians proach is proprietary. ists, someone will spend a are skilled in examining that year testing the information kind of data situation. There- COOK Report: You’ve made in the database according to fore we have to bring to bear it clear in your papers that some hypothesis. He or she the power of mathematics in the proprietary nature of the write a paper and submit it to developing predictive algo- software would preclude it. the Annals of Internal Medi- rithms for medicine. Under “Representing the pub- cine. Two or three years af- lic interest in health care”, ter that the paper gets pub- Multi-disciplinary research you wrote “CLEOS® is not a lished. Assuming it gets read cannot be conducted the way commercial product”. It is and assuming it’s right, it’s clinical research and epide- owned by the IDM Founda- still up to an individual physi- miology have been con- tion because the software can cian to integrate the material ducted. A multi-disciplinary be corrupted easily for nar- in that paper into his or her approach depends on building row commercial gain, as for practice. very large and standardized example recommending clinical databases that accrue products of a single company data on individual patients or always recommending COOK Report: What do you across time. But we also patent-protected medications mean by “real-time process- need the ability to integrate versus generics. ing”? data from millions of patients Zakim: What I mean by real- across time. This requires an “An equally important reason time processing, and what I extremely powerful computa- for non-commercial owner- interpreted Neggers’s mean- tional infrastructure. You ship of CLEOS lies in the gen- ing to be, and why it hit me, need very smart people to eration of otherwise unavail- is this. Miss Smith comes to build and maintain the data- able clinical databases. This your office. She suffers from base structure, to build a power, in commercial hands, diabetes and is age 30. With networ, and to integrate all of would privatize medical real-time processing, what- this information. I know that knowledge to make new ever entry Miss Smith adds to this is true. There is nothing medical knowledge a profit her fundamental database, in available now that can sup- center. And commercial her digitized medical records

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 71 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 within CLEOS, is compared COOK Report: That’s awe- COOK Report: Over time as with every other patient in some. you accumulate patients and the database in real time. data, the system should get The output decision as to how Zakim: That’s how it’s sup- more and more powerful. Miss Smith’s management posed to work. We call that should be changed or left the doctor making decisions Zakim: Absolutely. It gets alone - because of how Miss on the basis of clinical expe- more and more predictive. Smith’s parameters compare rience. But the literature to someone who had diabetes suggests the doctor’s clinical COOK Report: How you in- at 30 and... then had X event experience is defined by the tend to start this up over the at 40 or at 60. last patient he or she re- next 2 to 5 years within the members who had a similar context of an infrastructure It gets more and illness. such as the Netherlands’?

more predictive. Let’s say we have 50 million Zakim: First, I don’t know people in the database. Al- the details of what it is. Sec- So Miss Smith is compared most 8 million will have dia- ond, I’d have to know the with... let’s say we have 50 betes. Maybe 600,000 had details, not the nitty gritty, in million people in the data- diabetes at 30. So the data- a conceptual way. We’d need base. In real time, Miss base is cut to 600,000. All of to have a discussion between Smith’s clinical parameters that is done instantly. The those people like myself who are compared with 50 million system matches Miss Smith know about medicine, who other people in the database with all present and past pa- know what’s in CLEOS, who to determine what is the best tients whose data can help know what issues are in- way to manage Miss Smith the clinician recommend the volved in adding knowledge on January 26, 2010, at age best course of management to CLEOS. And I would want 30 with the following diagno- for Miss Smith’s diabetes, at a mathematician who has ses and co-morbid state, if that moment, in real time. been very involved in, for ex- she takes insulin, and this ample, analysis of micro- and that, and something from Do you know how many par- array data, which is closely hypertension, and has a cre- ticipants were in the Fram- analogous to the problem of atine of 3.6. The system rec- ingham Heart Study, which is analyzing the data CLEOS ommends a management de- the largest predictive data- collects. cision by the time Miss Smith base in our universe right finishes entering her data. now? It started with 8,000 COOK Report: I have the That is real time processing. and later they got 12,000 impression you’re starting more. And it’s an enormously with a few hundred people, Physicians must powerful database. But it then a few thousand people, does not have all of the data and now you are doing some be able to directly fields in it. And it doesn’t fol- data accumulation in Ger- low the patients as carefully input the knowl- many. And at this point it’s as an electronic system like focused on cardio-pulmonary edge themselves, CLEOS can. So imagine what data. Are you actually writ- you could do with a database ing code that takes the pa- or it will not work. of 600,000. tient questionnaire and follows it down the diagnostic

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 72 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 path where you begin to pull Oh, boy. How stupid could I COOK Report: Is it tree in specific medical literature? have been? I left something branching that’s not predict- out. The patient won’t notice able? Zakim: I don’t write code. I the sequence but it makes it have a graphic user inter- very difficult to ask follow-up Zakim: No, the trees are face. I write the code as I questions. So it’s a re- completely determined. How- add knowledge. Now this is a iterative process in which you ever, there are more path- very important point, be- enter knowledge, and you ways through the trees than cause you will not success- have to play the knowledge there are people on this fully build this sort of a sys- back to yourself to refine the planet. tem if you put knowledge en- knowledge. gineers between the system COOK Report: And the and the physicians. You will …not just very knowledge engineer is not go through hundreds of mil- large data sets, aware of the issue of those lions of dollars and you’ll pathways? come up with garbage. Physi- but enormous cians have got to be able to data sets… Zakim: Correct. He doesn’t work at an inter-face so that know where to put what. Not they directly input the knowl- only that, it will take too long Okay, now I work with a edge themselves. It will not to feed it back to him. And knowledge engineer, and I work otherwise. I know that. the ability to change things can tell the knowledge engi- on the fly and then see what neer I want him to program COOK Report: What would that did to it...that’s a very this set of questions in this be an example? powerful experience. I’ve sequence according to this learned that I can build logic set of answers. And then the Zakim: When you come to structures that I never see knowledge engineer sends my office and I ask you only when I’m building it. I only me an email saying, Okay, one question that is not re- see it later. you can see it. sponsive to something you’re telling me—which is: “What’s COOK Report: Is this one of COOK Report: I’m guessing wrong? Why have you come the example of the mathema- from what you’re saying, that today?” Now everything I ask ticians and mathematical al- there’s much tree branching subsequent to that is reac- gorithms? and perhaps fractals? tive. I interpret your answer in the context of everything Zakim: No. The application Zakim: There are some frac- you’ve told me, and then I of mathematics in Neggers’ tals in that I reuse nests of ask you the next question. view is the issue of multi- questions. For example, fe- disciplinary research, but also ver. Completely unspecific. It When I write this out, I’m the combination of mathe- comes up over and over working in an abstract situa- matical analysis with the again. It will be used in mul- tion. I have no responses; I computational infrastructure tiple places in a medical his- have to put in all possibilities. in order to provide real-time tory, in a million medical his- So I will write it and I will in- processing... not of just very tories, in more than a million corporate it. Then when I large data sets, but of enor- different places. play it back, which is looking mous data sets... worldwide at the interview, I can say, data sets.

COOK Report: So you have COOK Report: As the engi- NORDUnet articulation that already developed a pretty neers like to say, that is not follows. good questionnaire and pa- trivial. tient history form in your As one looks at the ongoing hospital in Stuttgart. But Zakim: I think it’s very hard need to fit the innovative when you want to take ac- to build an infrastructure that Dutch patterns to the larger cumulations of this patient absolutely meets the needs and more complex technol- data, and when you want to of any comer. But it appears ogy plans and roadmaps of complete more data in order the Dutch have probably built the European Union as a to do new kinds of clinical an infrastructure that makes whole, one begins to bump research that has been here- it relatively simple to meet into different planning phi- tofore impossible... that’s the needs of a variety of ap- losophies. What follows is a when you begin to need the plications. narrative summary of the ap- kind of system we’ve been proach that Kees Neggers talking about in the Nether- COOK Report: B o b outlined at a NORDUnet lands? Hertzberger’s whole approach meeting in September 2009. to building these e-science I let Kees paint his vision and then add my own commen- Zakim: That is correct. But stacks has been to make tary. The Powerpoint file it goes further than that. Be- them as modular and generic and video of the presentation cause every time data are as he can. The goal is that are available on the NOR- added to the database... when various disciplines DUnet website under speak- every time Miss Smith adds come in and want to add ers data: data to the database... her their own tools and their own http://www.nordu.net/confer data can potentially impact approaches, that they can do ence/ndn2009web/programm every other diabetic in the it as cost-effectively and e.html database. The knowledge of time-effectively as possible. what happened to her in re- Zakim: Even if they haven’t Neggers: I want to explain sponse to this or that (or gotten there yet, these are my perspective on the ap- what didn’t happen) could still very interesting people to proach to networking and impact everyone else in the talk to. These are the people knowledge infrastructure for database. who want to make it possible. research that we are taking in the Netherlands now. …it’s very hard to How to Organize the Engine to Let There are three trends in rebuild infrastructure Innovation search... that can meet the Flourish? First there is the emergence needs of any COOK Report: Kees Neggers of System level science which customer… and Bob Hertzberger are is defined as “the integration amongst the leaders in the of diverse sources of knowl- Netherlands articulating the edge about the constituent That knowledge needs to be research philosophy that parts of a complex system immediately available to the David Zakim found so fasci- with the goal of obtaining an next patient... who might log nating when I used Kees’ understanding of the sys- on in China. It’s another level tem's properties as a whole” of real-time processing. [Ian Foster].

Next is Multidisciplinary re- Other changes are leading to Our goal is to create an inte- search. Here each discipline new requirements for net- grated ICT infrastructure for can solve only part of a prob- works as infrastructure. We the Netherlands. In June2008 lem. Consequently you need have an explosion in the SURF made a proposal to the collaboration between differ- amount of data from experi- Dutch government to inte- ent research groups. These ments and simulations. Here grate existing organizations groups may be distributed are examples: LHC, LOFAR, for Research Network, Grid across states, countries, or e-VLBI, ITER, Climate model- Computing, Supercomputing, continents. ing, Humanities and Social and services for e-science Sciences. We also need near into a single organization re- You have Research driven by real-time processing of very sponsible for the ICT infra- (distributed) data. This hap- large datasets. It is neces- structure. And to finance the pens in the context of a data sary to be able to support an development of a single con- explosion, both in volume increase in remote collabora- trol plane and a consistent and complexity. You want tion associated with distrib- set of services for research both simulation and experi- uted sensors, shared comput- on a structural basis. ment combined. Your scien- ing and storage, grids and tists want to be able to ex- finally Virtual teams. In December2008 the Neth- plore data-sets with no up- erlands ICT Research and In- front hypothesis. We need a seamless, inte- novation Authority (ICTRegie) grated ICT infrastructure. An provided a similar recom- All of this requires cutting integrated ICT infrastructure mendation with SURF as um- edge network innovation for must facilitate the shared use brella organization. In May which internet is inadequate. of networks, computing 2009 the Dutch government For these needs a “Best ef- power, data storage, instru- agreed with the principles fort” network is not good ments, etc. It must ensure and asked SURF to work out enough. The Internet, as a hassle free end-to-end con- an implementation plan. best effort network was de- nectivity. It must be there as signed for delay tolerant, an infrastructure into which Implications for gov- many-to-many communica- like an electric utility users ernance models tion. can plug. It must provide a single user interface and a Governance models for the …a “best effort” single control plane for the ICT infrastructure need to allocation of multiple re- encourage innovation in or- network is not good sources, from multiple do- der to be able to offer ad- enough… mains and in multiple loca- vanced services without sac- tions. It must provide an in- rificing reliability. We also tegrated set of services to must ascertain how we can To meet our research needs, support all kinds of research. achieve all this. In my opin- the network will have to de- ion this should be done by liver guaranteed performance Necessary for the coordination and collabora- for large data flows and time- tion, in close relation with critical applications and do so Future - A National advanced users, not by creat- across multiple domains with Integrated Research ing monopolies different technology while Infrastructure keeping the successful end- Bottom up innovation to-end model of the internet.

- Innovation cannot be With a Decentralized govern- pared to the spill-over effects planned, designed and opti- ance model you get decen- of innovative services. mized tralized “unity in diversity” Hence, a centralized model - Innovation is like evolu- and the ability to experiment works well at EU(or larger) tion: unpredictable, but with with multiple solutions in level for very expensive one- a clear direction (at least in parallel. You get closer to off expenditures for large fa- hindsight) user needs and can facilitate cilities like LHC or ITER, not - Evolution needs three ele- organic growth. The price for pan- European communi- ments: will be some degree of dupli- cations infrastructures If 1. Variation: many attempts cation fostering a need for cost efficiency were the ar- to solve the same problem coordination. gument, we should all be us- 2. Selection: an environment ing the commercial opera- where the best solution is How to decide on the tors! given a chance to thrive best model... 3. Propagation: dissemination Why network innova- and replication of the best A centralized model is likely tion needs a decen- solutions [Source: Prof. Bart best for large-scale, single Nooteboom] tralized approach facilities (e.g. LHC). It also works for distributed, homo- Central organizations tend to The choice appears to be ei- geneous sets of resources become static. Why? More ther a Centralized govern- (e.g. LOFAR). On the other parties are involved in syn- ance model where on the ap- hand we think that the de- chronous decision making at parent plus side centralized centralized model is a good all levels leading to more “command and control” over option for distributed, het- talking and less action. Effort the entire infrastructure may erogeneous facilities (e.g. must be expended to ham- avoid duplication of efforts EGI, CLARIN, LIFEWATCH). mer people into a single or- and introduce economy of The best example is the ganization, single vision and scale. internet where you have a single approach. some coordination of com- Innovation is like munication standards and the …technology evolution: use of identifiers while everything else is decentralized. organization works unpredictable, but most effectively with a clear Is cost efﬁciency a valid argument? when it percolates direction bottom up… A centralized model may in- On the other hand one loses troduce economy of scale, In contrast decentralized col- variation and central planning but the costs of research laboration tends to create tends to stifle innovation. In networking are not deter- innovation. Smaller entities addition, in large communi- mined by the international can offer more flexible and ties decision making will be backbone network alone, but quicker response to a chang- slow and will lead to medioc- much more by the NREN and ing environment. More par- rity. campus networks. Moreover, economy of scale of a central ties working in parallel create approach is minimal com- different visions and “friendly

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 76 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 competition” of ideas within stantaneous communication Furthermore because the an agreed open interworking and collaboration fails to de- hardware delivers ever more environment. liver value. By lowering eco- bang for the same buck and nomic and trade barrier be- the e-science center has been COOK Report: As John tween member states, it can designed to commoditize the Seely Brown and John Hagel add value from the existence production of open source e- say in their collaborative blog of collective whole to the in- science middleware, Brussels Edge Perspectives in the 21st dividual members.. However has in this area only dimin- Century and in the internet as far as technology projects ishing returns to offer if the the only locus for creativity is go, given the fundamentals project does not rest on at the independent edge. of the network and how it some huge capital expense The center or the network enables the simultaneous de- like the LHC. There is a say- exits to provide a common centralized evolution of col- ing that all politics is local. To protocol by which the edges laboration, the time taken by it could be added the corol- may communicate with each getting the edges to agree to lary that technology organi- other. take directions from a pan zation works most effectively European, center-directed when it percolates bottom up It is here that the European technology plan will mean from the plans of people who Union formed to make a that the edges have lost from will pay for and use it. global economic power out of one to three years of time in the smaller member states in which they could have been the age of internet linked in- doing things on their own.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 77 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 VIII. How A Progressive ICT Infrastructure Beneﬁts the Economy The innovation engine as open fabric

Through a number of ad- a stimulus to economic activ- An important innovation vanced ICT infrastructure ity because the contribution driven by e-science methods projects, the Netherlands has to academic research has is a change in the research become an attractive country been sufficiently highlighted method itself. The infrastruc- for scientific research. The under Science Case. ture becomes an instrument, hub function of the Amster- delivering the flexibility to set dam Internet Exchange One of the innovative aspects up different partnerships as (AMS-IX), the global connec- of an advanced ICT research and when required. tivity achieved through infrastructure is the potential SURFnet, and the active par- to perform cutting edge re- Philips provides a good ex- ticipation by Dutch groups in search in a number of fields ample. In recent years Philips various international projects and consequently reducing Research has undergone a all have all contributed in at- R&D lead time. Through col- number of major changes, tracting high-tech companies laborative projects with in- refocusing from chips and and research centers. This dustry by academic research consumer products to health has led to a rich ecosystem of groups, an automatic transfer systems. Philips Research ICT related activities, as well of knowledge is realized. This now finds itself with a differ- as research activities enabled mechanism and the results ent set of priorities and needs through the ICT infrastruc- from VL-e, MultimediaN, to work on subjects such as ture. Examples of this type of NBIC, BiG Grid and GigaPort biomarker discovery where concentrated activity can be have already led to extensive there is far less expertise seen in the High Tech Cam- collaborative efforts with ma- than in the traditional areas. pus in Eindhoven and in the jor industrial partners includ- Amsterdam region. ing Philips Research, Organon To accommodate this focus (then part of Schering-Plough shift, Philips Research has The concept of e-science as now part of Merck), DSM, adopted the open innovation depicted in the science cases and Logica/CMG. concept, which has resulted shown in the preceding chap- in a more open organization, ter is innovative by itself: …new forms of (especially for a company completely new forms of re- founded in 1891.) It stimu- search are possible with the innovation lates strategic cooperation availability of an ICT research become possible with other companies, uni- infrastructure that provides versities and research insti- state-of-art computing, stor- with state-of-art tutes. Facilities are shared age and communications, as computing, storage, with partners in order to im- well as instrumentation for prove cost-effectiveness and the researcher. In this sec- communications and efficiency. The Research ICT tion we focus on the innova- instruments... Department has modified its tive aspects of the infrastruc- strategy in order to enjoy full ture on industrial R&D and as advantage from this change.

Research networks and e-Infrastructure for Research in general act as innovation agents. The need to serve a demanding user community of leading edge researchers forces e-Infrastructure providers to invest in pilot services ahead of the market. In doing so they create an innovation engine that works in many dimensions: it creates a demand pull for new networking research and products and a market push for new telecommunications services. At the same time it triggers development of new middleware and applications, which then themselves function as engines for innovation in their layers. As a result it not only shortens the development cycle for new services, in many cases it allows services to mature which would never be able to do so on a purely commercial basis as is nicely explained in the interview with Hans Dijkman later in this report.

Through collaboration with While the 'Philips' case is al- search. Similarly, Organon GigaPort, VL-e, and BiG Grid ready very convincing, more has moved its main focus (for which Philips Research industrial involvement is on from bio-informatics to e- will host a significant part of the way. In the NBIC consor- Bioscience, allowing it to the infrastructure), the de- tium several industrial part- adapt quickly to new topics in partment is now able to pro- ners (e.g. Keygene) have drug design. vide networking and comput- also expressed their intention ing facilities to the research to become partner in the dis- Building an ICT infra- groups - an action on the tributed ICT research infrastructure for research part of Philips Research which structure, starting with hous- would not have been feasible ing a 'life science cluster'. creates spin-off for in any other way. This will closely follow the industry model chosen for Philips Re- Not only does the ICT re-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 79 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 search infrastructure enable pioneer better customer sup- ers to push the envelope in innovative research in the port strategies far ahead of advanced networking. Or- larger industrial community. the market. In doing so they ganizations connected to Building and evolving such an create a fabric-like innovation SURFnet have been encour- infrastructure itself is a engine that works in many aged to upgrade their own stimulating formidable chal- dimensions: it creates a de- internal infrastructures to lenge in itself. It creates mand pull for new networking make the best possible use of many opportunities for inno- research and products and a available services. Vendors vative activities. This will un- market push for new tele- were encouraged to imple- doubtedly stimulate regional communications services. ment the newest technology and national economic activ- And, as has happened here, in their offerings. ity. it triggers development of new middleware and applica- Service providers have been A parallel can be seen with tions, which then themselves able to use SURFnet as a the development of the function as engines for inno- platform to experiment with Internet and the World Wide vation in their layers. This innovative services for so- Web, which – in the Nether- shortens the development phisticated user communities lands – started largely in and cycle for new services, and in and early adopters, before around scientific institutions. many cases it allows services rolling out to the wider popu- This has led to the emer- to mature that would never lation. As a result, not only gence of a great variety of have been able to survive in the research community but ICT related business activities purely commercial market- the ICT community as a – as has been demonstrated places. (This process is de- whole has benefited from the in other countries as well. fined and explained more innovation initiated in the Gi- The ISP industry (Internet fully in a conversation with gaPort projects. Service Providers) grew out Hans Dijkman in the next of the slipstream of the de- chapter, “SURF as Economic The picture on page 71 above velopment of the Internet, Midwife for Technology illustrates the ‘innovation en- which has been boosted by Transfer”.) gine’ effects of GigaPort. SURFnet, it is more than likely that the development of Innovation is conta- Part of this is a general en- generic services for research gious thusiasm for progressive use will stimulate the emergence of new technologies. There is of services-based industries, also an almost patriotic rec- Both the research institutes offering a new generation of ognition that this is supposed connected to the infrastruc- opportunities and economic to grow the economy and im- ture and the vendors supply- models. prove the quality of business ing the various components offerings. It’s in the unwritten to ICT researchers and sup- Research networks and e- job description that every- port teams are themselves Infrastructure for Research body involved with evolving often inspired to innovate often serve as innovation technology is also involved in within their own environ- agents. The need to serve a new businesses, new prod- ment. SURFnet is a good ex- demanding and knowledge- ucts, new employment and ample. In the GigaPort and able user community of lead- new revenue. GigaPort Next Generation ing edge researchers forces projects, SURFnet has chal- e-Infrastructure providers to At it’s most productive, the lenged both vendors and us- invest in pilot services and to productivity and value-

An illustration of the Netherlands “e-infrastructure” from Bob Hertzberger’s VL-e project.

A diagramatic view of what in the United States we call “cyber infrastructure”. The VL-e picture shows the multidisciplinary collaborative use of the infrastructure, while GLORIAD picture shows the data ﬂow resulting from worldwide collaboration. For more information see the Gloriad web site http://www.gloriad.org/gloriaddrupal/

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 81 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 generation that Kees Neggers SURFnet has presence in networking or- and Bob Hertzberger have ganizations as GEANT, GLIF described on the pages of challenged vendors and TERENA, in HPC-oriented this document is what Carlota and users to communities such as DEISA Perez would call a case of and PRACE, and in grid- radical centralism. In the push the envelope oriented communities such as words of Harvard Professor EGEE and EGI. [Editor’s and sardonic songster Tom Note: Cees de Laat noted Many scientific communities Lehrer, (a favorite of a earlier that significant leadership in the Netherlands are using generation of geeks), it’s these national infrastructure roles and contributions are about “doing well by doing also done to IETF, OGF and facilities as a basis for their good.” W3C. And Bob Hertzberger collaboration. For example, as another illustration adds the SURFnet6 network was that in mid 2009 Nikhef won Partnership Case created by a broad consor- tium of academic and indus- the bid for the European Grid Infrastructure (EGI) head of- try partners, for use by the By definition, an ICT Re- fice.] The innovative light- search Infrastructure on a entire research community. path exchange NetherLight national level is a shared The ICT infrastructure pro- has grown into an important concept. This does not only vided through BiG Grid will be hub for networking connec- hold for the collection of used by the Nikhef high en- tions, leading to new partner- ergy physics community hardware resources and ge- ships both at the network neric software development, (linked to LHC at CERN), by and the applications level. it also holds for the individual the astronomy community resources themselves: the (LOFAR), by the life sciences network, computing systems community as represented by Technical case NBIC, by alpha science com- and storage facilities. For General description munities like DANS, and by years, scientific communities The science and innovation computational sciences com- in the Netherlands have col- cases described previously munities represented by laborated in realizing large- require networks, computing, NCF.” scale facilities as a superior and storage technology. Link- research network and large ing these together in a co- high performance computing an almost patriotic herent and more seamless systems. Collaboration in this way, using web and grid way has not only led to recognition that technology and employing e- economies of scale, but has this is supposed to science methodologies, is a also led to new styles of mul- condition sine qua non for tidisciplinary scientific work, grow the economy… progress in almost all scien- as a result of using large ICT tific fields. Thus a landscape research facilities together. Due to its active attitude to- of resources must be formed, Furthermore, it appears that wards international coopera- ranging from networking, ca- in The Netherlands, both tion, and to its advanced in- pability and capacity based academic and industry re- frastructure, the Netherlands computing resources (see search groups are involved in is in a unique position to at- under computing and storage this type of collaboration. tract international collabora- resources), storage and data tions as well. This has al- services, to application sup- ready led to significant Dutch port services.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 82 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 …cutting edge high contained in the highlighted the network. These proper- box, which ranges from the ties make them very well performance network to generic e-science suited for transport of large networks are indis- services and the tools to de- quantities of data, such as liver them. originates from scientific ex- pensable for periments, or for high per- The Network formance interconnection of modern research… grid storage and computing The availability of cutting- components. snip edge high performance net- The total chain of resources, works is indispensable to Computing and middleware and services de- modern research, as it con- termines the added value of nects distributed scientists, Storage Resources the infrastructure for the re- instruments, computing and Capability-based Comput- searchers. The people who storage facilities. These net- ing become skilled in working in works are no longer distinct such an environment are a from the rest of the ICT in- Different problems have dif- resource in their own right. frastructure: resources within ferent computing needs. Allo- The human actors in a pro- the network and within the cating the most efficient gressive infrastructure can be rest of the infrastructure will computational resources to recognized as becoming in- have to be controlled through problem areas not only has valuable “components” of the an integrated middleware scientific advantages (only infrastructure as well. It is layer, to ensure an optimal those groups in need of cer- often these infrastructural allocation of resources. tain high end high cost hard- actors who choose to break ware will gain access, off and start a new business thereby increasing effi- or create a new innovation ICTRegie ciency), but also has budget community. facilitates the advantages. This has long This infrastructure can only been recognized in The Neth- be realized by developing the formation of erlands, leading to so - called necessary services through innovation “capability resources” and an integrated effort – and “capacity resources”. with a commitment to open platforms technology. This is what as- ICTRegie and the sures opportunity to re-use SURFnet6, the current gen- the data processing services eration of the Dutch national Dutch ICT and thereby the re-use of research network, is a state- Infrastructure knowledge. More value is of-the-art network and is thus created. widely recognized as a world ICTRegie, the Netherlands leading network. SURFnet6 is ICT Research and Innovation A layered approach a hybrid network, providing Authority, as its web pages both IP connectivity and explain: “is a compact, inde- Schematically, the infrastruc- lightpaths within the same pendent organization consist- ture model is represented by network. Lightpaths are dedi- ing of a supervisory board, the figure below. The techni- cated high-capacity, low la- an advisory council, a direc- cal case addresses the ICT tency point-to-point connector and an office. The Minis- research infrastructure, as tions between two nodes in ters of Economic Affairs, and

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 83 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 of Education, Culture and articulation which means forms, parties like knowledge Science bear the political re- stimulating and keeping track institutes, ICT businesses, sponsibility for ICTRegie. of (latent) demands from and representatives of The organization is supported customers and businesses. potential users or customers by the Netherlands Organiza- ICTRegie encourages this ac- come together to explore tion for Scientific Research tivity by facilitating the for- meaningful utilization of ICT (NWO) and SenterNovem. mation of ICT Innovation and collectively formulate platforms for different areas topics for further research.” ICTRegie effectively gathers of expertise (e.g. Health http://www.ictregie.nl/ictregi together the stake holders for Support, Product Software). ehome-FAQs.html . ICT. It engages in demand Within these Innovation plat-

Above left: Kees Neggers, Cees de Laat and Gordon Cook and above right Bob Hertzberger, Hans Dijkman, Cees de Laat and Gordon Cook at Supercomputing 09 in Portland Oregon.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 84 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 IX. SURF as an Economic “Midwife” for Technology Transfer An interview with Hans Dijkman, Kees Neggers and Bob Hertzberger

Neggers: Meanwhile Hans The functional composition sponsible for making generic Dijkman has some slides ex- diagram below illustrates the e-science tools or middleware plaining how the SURF inno- e-science infrastructure that that can be used by the spe- vation model fits into the produces the research that cific scientific domains. The larger economy as a whole. we are talking about now. activity is not limited to only engineering (i.e. a form of Dijkman: Let me talk about We are putting an e-science applied science) but also re- what I will call “Product De- Research Center between the search is done to develop velopment 101.” Over time ‘hard’ infrastructure and the new scientific methods and research is done with the aim specific science domains. The techniques that are made of introducing new products. center is, for example, re- possible with modern ICT.

Functional composition of the E-science Research Center and ICT infrastructure for scientiﬁc research. Source is ICT Regie.

which we call the starting point. Here we have fundamental research and it is unknown whether there will be market pull or market push. But over time some demand can emerge. People see it and say- “oh I could use that.” They become the first users from the general domain in which the research is done. This is the first phase of demand. When you have enough demand at vertical axis A, supply begins to be The classical Product Life Cycle. The different stages of offered. the ‘life’ of a product in time.

At vertical axis C, supply and Applied and fundamental sci- the innovation stage we ex- demand meet in a plateau ences interact with each panded the model. and a competitive market for other in the center. In our the new product emerges. vision e-science stands for …three market stages: ‘enhanced’ sciences. That basic research… On the left side – axis A - the also implies that the interac- product is purely new and tion is collaborative. Different non-competitive market… innovative, but in the middle domains can share each and competitive market… as other people begin to want other’s resources independ- it, you have a non- ent of time and place. For simplicity we assume that every product can be traced competitive market beginning to develop. At vertical axis For financing we used a spe- back to some basic research B, standards are beginning to cial approach. To illustrate this we use an expanded version of the familiar Product Life Cycle (PLC) model to explain the need for an e- Science Research Center and the need for governmental funding.

We all are familiar with the theory of the PLC, usually presented as the following This model does not represent the product development stage or the preceeding innovation stage. To describe The ‘life’ of a product in terms of its maturity in time. From ‘birth’ to a standardized commodity. The demand and supply curves are –delayed- following the product life.

companies, for example Phil- ips and AT&T, are able to cover the cycle from basic research to the market. For them this middle stage is used e.g. to obtain patents and the IPR. In this way they get a competitive advantage. Let’s call this Proprietary In- novation. Other companies are starting from existing technologies, in other words, when the academic R&D (sub)stage is finished. In the case of e-science, the goal is to develop new methods and The ‘life’ of a product in terms of its maturity in time. But now added the techniques by which we do ‘drivers’ of the different stages. science. E-science leads to Open Innovation. This means emerge and this emergence isting technology, which aca- that, under certain condi- enables the increase in sup- demic scientists would nor- tions, ideas, results and ply to develop into a competi- mally describe as develop- products are shared. The fi- tive market. Altogether, we ment. This different use of nancing should reflect this distinguish three stages; ba- the word “research” can lead character. sic research, non-competitive to many misunderstandings. market and the competitive If we represent the middle The output/results of basic market. stage as follows, [see non- research is rarely ready to competitive market time market. The quality require- The output chart below] the problem be- ments are completely differ- comes more clear. Only a few ent. So, in this stage, the re- of basic research is rarely ready to market.

What are the driving forces in the different stages? Basic research produces publications which are what keeps people at universities alive. In the competitive market is it simple: money. But what about the middle stage? In industry the term “research” is frequently used to describe innovation with ex- The non-competitive market stage of a product but now in more detail.

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 87 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 sults from basic research have to be upgraded to a level so that others can use them. Prototypes have to be engineered into production types, computer applications have to be made more user friendly, and so on. An important part is also standardization. For e-science, it is the stage where applications are made generic and methods are developed to give access to different resources.

It is a necessary stage but pull. At this point Venture with no clear incentive or re- Capitalists show up in case of Different driving forces result. It’s not where you spent potential commercial value. quire different organizational your basic research money. In the case of e-science this models. Therefore it is nearly It is also not where compa- is not working. It is here impossible to put them under nies invest for their product where additional governmen- one management. If we focus development. This “in be- tal sponsorship is needed. on the e-science research, we tween kind” of stage needs Otherwise nothing will de- have made a choice for dif- an external force, a push or a velop. ferent organizations for different stages.

So SURF here in the blue section immediately below provides the bridge of support between the early adopters and the arrival of a competitive market. The early adopters are the pioneer users you need in order to mature your products and services. You make something. Some- one likes it and starts to use it and then you get the demand for help. Users want a help desk. Using your grid library needs other forms of Regardless of the nature of the product, humanities, behavioral sciences or support. Until it becomes a science, the maturity is about the same. In the e-Science Research Center they commodity that you can buy can come together as building blocks for an e-science application. It also shows in which stages VL-e was operating. in the marketplace, you need

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 88 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 a source of support for this. COOK Report: OCW is? larger and permanent – not But also engineering is just a project. needed to transform e.g. an Neggers: The Ministry of application from prototype Education, Culture and Sci- Hertzberger: Correct. into a more robust, user ence. This Ministry gives friendly, etc application. money to SURF and NWO – Neggers: This is the glue not the ministry decides that the government has to This is the glue which researchers get what inject to make this shift from money. NWO is going to fi- research to the market place. the government has to nance this multidisciplinary e- While it also stimulates multi- inject… Science Research Center disciplinary applications for to shift from (eSRC). And based on the the first time. research to market… ICTRegie advice the ministry is assumed to fund the blue Hertzberger: You can never pilot phase under the SURF get these guys to do this Above is where the different umbrella via SURFnet and themselves. There is no in- scientists are funded by the SURFcdf. SURFcdf is the centive and they lack the ex- Dutch science foundation – working title of the still to be perience. You need multidis- the NWO. The three research financed High Performance ciplinary interaction between domains are alpha – humani- Computing and Data Facility. computer scientists and those ties, beta - hard sciences, domain scientists. You need and gamma - behavioral sci- SURFcdf is a work in pro- it but don’t have it and con- ences. gress. The intention is that sequently you setup an ac- the money spent on super- celerator to leverage existing COOK Report: The above computers and on grid and applications and develop new VLE triangle is the develop- data storage projects are kinds of applications. ment investment flow? merged into a single facility making it available to all dis- Neggers: But you also need Neggers: The important ciplines. to underwrite the computer thing to note here is that the science part of the entire red blue and green are nor- COOK Report: So this is middleware that is needed to mally vertical pillars doing how you have institutional- permit progress along this things entirely on their own. ized the support necessary to path to establishing this new Through VL-e (Virtual Labo- bring basic research to mar- paradigm for interdisciplinary ratory for e-Science) the ket? e-science learning. It has black boxes representing the both a computer science and three broad types of research Neggers: Bob, correct me if an application component. It cut across the otherwise in- I am wrong but your VLE pro- is the merger of the two that dependent agencies and ject was successful and now creates the added value by make interdisciplinary re- ICTRegie has advised the leaving the generic part to search possible. There are government that when this the computer scientist and some tools that you need project ends it should be- the application part to the within single research do- come a new organization not domain scientist. Together mains but other tools are just for the few application they have to make something needed as well that cross sci- people that you were able to that works. You need them entific domains. serve with your tiny bit of both. money, but it should be

Dijkman: Correct and if you descriptions of the innovation open environment. No IPR, do it right you make way for processes of Xerox, IBM, Lu- no patents, etc. I also think what Henry William Ches- cent, Intel, Merck, and Mil- that the outcome of e- brough was talking about in lennium, and the many spin- science, e.g. generic tools to his 2003 book Open Innova- offs that have emerged from support applications should tion: The new imperative for these firms. Open Innovation be handled like Open Source creating and profiting from shows how companies can Software. technology. I quote: use their business model to identify a more enlightened Most Webservers are based "In today's information-rich role for R&D in a world of on open source software environment, companies can abundant information, better called 'Apache'. The core of no longer afford to rely en- manage and access intellec- Apache is maintained and tirely on their own ideas to tual property, advance their managed by the Apache advance their business, nor current business, and grow foundation. They protect the can they restrict their innova- their future business. Arguing quality, etc. Everybody else is tions to a single path to mar- that companies in all indus- not allowed to change the ket. As a result, says Harvard tries must transform the way core but is allowed to build Business School professor they commercialize knowl- add-ons. There are more ex- Henry W. Chesbrough, the edge, Chesbrough convinc- amples. You can imagine traditional model for innova- ingly shows how open inno- that something equivalent to tion - which has been largely vation can unlock the latent the Apache foundation is internally focused, closed off economic value in a com- formed. This 'foundation' de- from outside ideas and tech- pany's ideas and technolo- velops and maintains a set of nologies - is becoming obso- gies." e-science applications, etc. lete. Emerging in its place is But e-science is also sharing a new paradigm, 'open inno- e-science of e.g. databases. Between vation', which strategically academic institutions no leverages internal and exter- needs to happen problem but sharing data be- nal sources of ideas and in an open tween companies becomes a takes them to market lawyer’s paradise. A spin-off through multiple paths. This environment… of e-science should be, in my path-breaking analysis is vision, a kind of Clearing based on extensive field re- house. A place where data I have talked about 'open in- search, academic study, and can be shared without the novation' versus proprietary the author's own longtime innovation. I believe that e- legal bureaucracy. experience working in Silicon science must happen in an Valley. [The book offers] rich

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 90 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 X. Coming to Conclusions Realizing the beneﬁts of a re-usable infrastructure

COOK Report: Lessons frontiers. The future will be- The earliest perception of learned in the Netherlands long to those countries and blogs as reported in the me- should be an education for communities who are proac- dia – that they were the the rest of the world as well. tive in investing in national mindless daily diaries of mil- It is a testament to the power infrastructures that serve the lions was faulty. Of course of the commons — and to the many rather than the few. there was some truth to this willingness to learn how to Over time the world will see but what really mattered was work together. It is about the that infrastructures left in that blogs enabled the form- benefits the Netherlands has walled gardens will wither. ing of community of interest demonstrated can result and joint problem solving. In from sharing the network, Lessons learned in effect, they began to “grow” the knowledge, and the tools. fresh new commons that in- And it is that “spill-over” ef- the Netherlands creasingly inter-connected. fect Kees Neggers of SURFnet should be an And that takes us back to e- and Wim Liebrand of SURF science in the Netherlands. keep mentioning — in which education for the the benefits of creating a na- As discussed earlier in Chap- tional knowledge infrastruc- world ter V “Making e-Science ture are widely distributed Work” there are at least two and pay back results which of An area where it is easy to vectors of differentiation for exceed expections. recognize the growing oppor- e-science being pioneered by tunity enabled by open infra- important e-science initia- At the same time, we have in structures is e-science. With tives like VL-e. comparison the rest of the e-science it is “early days”. world — the United States The temptation is to compare The first is supporting a included – where so many it to where we were in 2000- methodology for science that have chosen to focus on pri- 2001 with blogs and wiki’s. embraces trans-sectoral and vate not public good. Where People were experimenting the consequence has been with both, but neither the with limited funds one of short term horizons tools nor the culture were in rather than long term in- place to enable really sys- decision-makers vestment. That needs to be temic widespread use. We had better put it in changed, or at the very least needed another six years be- moderated. fore the possibility for global the infrastructure collaboration was manifest The power of networked and could be articulated by first technologies to change the Yochai Benkler in his book rules going forward cannot be The Wealth of Networks. multi-disciplinary collabora- denied. As the physical world http://en.wikipedia.org/wiki/ tion. The second is creating runs short of opportunities, Yochai_Benkler. generic middleware tools that the digital world offers new suddenly make it possible for

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 91 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 different disciplines and much as possible at layer 1 ginning to enabling the communities to begin col- and layer 2 and going to equivalent of the blog and laborating. three only when absolutely wiki for global science. And necessary. It collaborated as the Fourth Paradigm book The future will be- with the likes Steve Goldstein has shown, there is already and Tom DeFanti and started tremendously impressive long to countries slinging 10 Gb lightwaves work going on globally in from Amsterdam to Chicago. data intensive science. and communities This was in 2002-2003. But most proactive immediately Cees de Laat But this must be done in par- found it necessary to adjust allel. Technology never in investing the behavior of TCP/IP in sleeps. Optical development in national 2002-2003. (Doing real infra- hasn’t stopped in Nortel's structure is not for wimps.) common photonic layer. It infrastructures. That allowed SURFnet to do keeps on getting better. things demonstrate the abil- Anyway, bringing all this to- Science in the past rewarded ity to conduct real-time four gether in the application layer the expertise of the “hedge- times high-definition telep- has stimulated the begin- hog” who was totally focused resence sessions in Japan nings of a truly global user on a narrow specialty and and North America and science movement where encouraged to stay in their Europe. people in quite a few other silo. Today in a world of countries besides the Nether- growing ubiquity of connec- ….adjust behavior lands are beginning to ex- tivity, the hedgehog will find periment. himself useless, unless he of TCP/IP to can figure out how to weave optical network In the process of preparing himself into a larger fabric. t h i s d o c u m e n t , B o b Increasing, as Cees de Laat properties. Hertzberger, Director of the said earlier in this volume, Virtual Laboratory for e- science is now global. To get Now, at the same time, and Science, wrote the COOK Re- port that “for e-science to at real problems and gener- in parallel with its FES infra- work a huge effort has to be ate real results, science must structure calls, the Nether- put into innovation. But tech- go where-ever the knowledge lands team is working on de- nology developments deter- trail takes it - must cross fining what the next level will geographic borders, and be. It has pioneered the next mine the pay back to customers.” This stimulated ad- knowledge domains as well. highly important integrative ditional conversation with What does that require in stab of rethinking the net- Hertzberger. terms of infrastructure. work stack up to the application layer. The application SURFnet laid the basic utility layer is the rubber meets the Hertzberger: We have to put tremendous effort into grid and the basic foundation road for e-science users. It’s innovation before we can il- on which everything else is where user connectable lustrate that it pays back to built. Then the team consid- lightpaths and Web services, our customers. How long, ered basic progress in optical remote instrumentation and networks and understood the grid and preliminary e- and more important, at what rate we absolutely do not cost effectiveness of doing as science middleware are be- know, because it depends on

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 92 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 the speed with what technol- instrumentation as well as ogy around you develops. the fact that you have to in- Like it or not this requires The only thing you know for vest a large number of years changes in the sociology of sure is that when you don't before you get your return on science. Specialized young do it you will loose. that investment. scientists had better understand that e-science - as a For e-science to However as politicians are form of instrumentation - will not scientist or entrepreneurs irreversibly change their pro- work a huge effort you seem to have to promise fession, and that because has to be put in that the sky is the limit be- science is conservative they fore they believe (and please better start today rather than innovation, but notice I don't say under- tomorrow. That in order to stand) that what we are do- realize its advantages they technology ing has some relevance and have to accept globalization developments must to be funded. in their domain and that a consequence is that they decide the pay COOK Report: This is called have to work in more multi- back to customers. building infrastructure and disciplinary teams. They have you do not get payback from to realize that the data/ infrastructure in 90 days and information centric angle – Consequently what you do is maybe not even in 900 days being collaborative - is often try to take the competitive but you do get incremental considerably different from advantage. We have had payback during the life of the the way they have learned to some visionaries (especially infrastructure. operate up till now. Hans Rosenberg) who saw early on the importance of Science globaliza- COOK Report: But your e- networking for a small highly science infrastructure build- developed country. The e- tion demands for ing process needs money to science case has been made education towards keep it going. Consequently possible by the existence of it is necessary to explain to an excellent network and the multi-disciplinary the funding agencies how it fact that science, in particular works and how it relates to medical science, is well de- scientists developments in other coun- veloped in the Netherlands. tries in the world and how Hertzberger: Yes and note one may most accurately But most important of all again a scientist is educated think about the dividends that, as a young scientist, I to solve a problem in his of- that one can expect to see. had the privilege of working ten mono disciplinary field. But such dividends that can- in an experiment which got However, at the same time, not be mature in 90 days or the attention of the whole because of the potential of 180 days without the basic world (even the New York networking, science is also foundations needed for global times) and for which the becoming a global activity. collaborative scientific inquiry leader got the Nobel prize That consequently e-science and even more general learn- [the discovery of the W and Z has to support that globaliza- ing bosons in high energy phys- tion first before the end user ics]. From that experiment I can obtain full advantage of Hertzberger: I would say it learned the importance of e-science for his applications. this way. That for the gov-

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 93 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 ernment and the decision structure. That is the layered makers the problem is that COOK Report: And the model we use in VL-e. It is they do not have unlimited point is, as the Netherlands certainly not complete and funding and that conse- has shown time and time requires a lot more research. quently they had better first again, if you don't do it first put it in the infrastructure. others will. There is a prime COOK Report: And it is However, because this is a mover advantage. It is defi- clear that you are thinking new game, established scien- nitely better to be building very creatively about how to tific communities all will claim the third or fourth story of follow existing ideas that they can do it far better in your building while your have been used in software their field of science if they neighbors are only complet- development to make operat- don’t have to spend time on ing the second floor. And if ing systems more easily all these new things because you use generic components compatible with continuing it only slows them down. like you used the saw mill in improvements in microproc- There is a continuing tension the 17th century you can build essor technology. Namely between the old ways and new floors on your edifice you could never afford to de- the new. Consequently there faster than the others. velop the science silos inde- is a huge risk that every field pendently for each scientific is re-discovering the wheel Hertzberger: Yes. in a discipline. You have to think over and over again and that modest way building up a re- collaboratively and to think you have to prevent that by usable infrastructure is ex- about the processes that are adopting generic solutions actly what we try to do in the needed in each discipline that and re-use of components. VL-e project. And then I can be generalized to a plat- have the vision that similar form that is usable between …mono-disciplinary ideas like the ISO/OSI model disciplines. This is a high- which was designed as a ref- level meaning of the process attitude results in erence model for networking that you've been explaining. and the virtual machine Furthermore you find that it huge risk in model that has been used to also has added value within a rediscovering the avoid that for each new mi- discipline because it provides croprocessor you had to de- rationalization of parts of the wheel over and sign a complete new compiler research process itself as a over again... chain might also work for form of pre standardization. part of that e-science infra-

If you want to know how far And in the same way, when time. Winston Churchill once the United States still has to people think about a national described war as being a cir- go in terms of re-thinking, ICT technology infrastructure cumstance in which “all re-designing and actually they like to think about things are on the move si- building a national infrastruc- hardware, software, net- multaneously.” What Moore’s ture for the 21st Century, a works, storage, servers and Law and Metcalfe’s Law and good place to start is databases. maybe Renan’s Law, too, all http://www.infrastructurerep mean is that we no longer ortcard.org/. This website is 19 years ago, when it was need war for the experience. run by the American Society first published, the COOK Re- Now all things are always on of Civil Engineers to track the port focused on technology. the move simultaneously. decline of America’s crum- First the bubble burst in bling infrastructure. Every 2000, and then the World The genius of the Dutch is year Trade Centers came down on that they chose to focus and infrastructurereportcard.org 9/11 in 2001. In both cases, invest in the one kind of na- publishes an annual grade for failures related to technology tional infrastructure which each of the critical infrastruc- were directly involved. There would assure them maximum ture sectors that make up was more to technology than opportunity and competitive infrastructure for the United technology, so to speak. By advantage for the Nether- States. Then it averages 2003 the COOK Report cov- lands, all who lived in it, those to assign an overall ered Technology, Economics learned in it and did business grade rating. and Policy, and the disruption in it. In 1997 they embarked and chaos which seemed to on the challenging task of The bad news is that in 2009, ripple around that combina- building a national knowledge the American Society of Civil tion. It was clear the Internet infrastructure supported by Engineers gave the U.S. in- and the growing fabric of the most participatory and frastructure a “D”. Not quite services it carried were rap- progressive ICT Research failing, but almost. When it idly becoming the new infra- technology network, we comes to Information Com- structure for the 21st Cen- think, in the world. The pre- munications Technology in- tury. Unpacking that became viously ten chapters have frastructure the news is even the core activity for the documented the extraordi- worse. There’s no mention of COOK Report. nary momentum they’ve cre- ICT or the Internet at all. ated. Given we live in a world (But America did get a C+ in Most decision-makers prefer in which atoms are limited, solid waste.) to think in terms of a world but bits, evidently, are not, it they can control, but the was an excellent strategy, Engineers still like to think world we face is one of con- and many now view them as about national infrastructure tinually accelerating change being the poster child for in terms of bricks, mortar, in which multiple paradigms digital infrastructures. bridges and… solid waste. are changing at the same

The individual who seems to “A golden age of complished at all stages and understand best the impor- at every level. worldwide expansion tance of technology for national development is Carlota is possible. Most importantly — and Perez, and a favorite of the Making it happen will something that is rarely car- COOK Report. We devoted ried out by other countries — our entire May issue in 2008 require thinking big, it must be negotiated with all to her work. She is both a deciding wisely and stakeholders, again and visionary and a realist. acting boldly.” again, over many years. Peo- ple focus on the technical ex- She titled a recent talk “The cellence of the new Dutch Future is Not Predictable Yet Carlotta Perez ecosystem. But the real gen- We Can Learn from History.” Epilogue: The World ius is in the inclusiveness of This, of course, defines the the Dutch process, the care- at the Turning Point challenge of implementing a fulness of the governance, successful infrastructure for Technological Revolu- the attention to policy, exe- the 21st Century. Much of tions and Financial cution… and total account- what information technology Capital ability. does is find ways to make the past work with the future. So at this point, given suc- And as change accelerates, It makes it seem as if this is cess in the Netherlands, but a so must agility. It takes eve- a single decision, a single act, certain amount of confusion rybody involved to optimize making the investment, etc. in the U.S., the logical ques- the process. The Netherlands In other words, she describes tion is why cannot the U.S. has proceeded by means of — and governments and just use the models? an inclusive cooperative those who serve in them tend to think the same way —it as process. A History Lesson from a matter of big dramatic The question is the same for steps. the Tsar of Russia, every nation, big or small, Peter the Great poor or rich, developing or But what we’ve learned from mature or in decline. How the Netherlands, in doing this As Carlota Perez promised, can we negotiate the transi- special report and an earlier history does offer guidance. tion from the past to the fu- report last year, contradicts Because this is not the first ture? As a community? As a this. It is more than a matter time the world has looked to nation? And increasingly, as a of big steps. In fact, imple- the Netherlands for help in species. menting a technology-based understanding technology technology infrastructure is a and infrastructure. In the This is what Carlota Perez very complex and long proc- 1600s and 1700s, when na- thinks about, talks about and ess that requires truly the tional infrastructure was de- writes about. Here is some- best of intentions, but also fined in terms of roads and thing she wrote in 2002: considerable realism, prag- canals, and no one in the matism, patience and persis- world matched the Dutch for tence. And we should not ex- understanding canals, water clude the virtues of balancing and the first brilliant efforts consistency and flexibility ac- at environmental engineering.

So there is much the U.S. Once upon a time in the U.S. That is why, in 1697, Peter and the world have to learn we respected expertise and the Great, enormously tal- from this new 21st Century knowledgeable technology ented Tsar of Russia, jour- knowledge infrastructure. But counsel: namely, The US neyed west to Amsterdam we have to make certain we Congress Office of Technol- and Zaandam to learn Dutch pay attention to the whole ogy Assessment, known as technology. He studied with and not simply the shiniest the OTA. shipwrights and experts in parts. many fields, to learn technol- When Newt Gingrich abol- ogy necessary for a national We need to begin with recog- ished the OTA in 1994 he infrastructure, the building of nizing that the Dutch have a abolished the US equivalent ships and fortresses and how thoughtful, complete and of the Committee of Wise- to design locks for canals. balanced governance mode men. He opened the field to which takes into account the the K street Astroturf organi- He returned to Russia with as needs of the private sector as zations that have made sure much technology knowledge well as the public. And it is a that US politicians remain as he could carry. This was a very inclusive from another ignorant of what goes on significant opportunity for point of view. abroad. I worked full time Russia. But unfortunately he from September 1 1990 did not bring with the thing Kees Neggers insists it is the through March 1 1992 on my that made Dutch technology users who run SURFnet. He own project at OTA. I can successful. He did not bring likes to say: “all politics is speak from first hand experi- to Russia the Dutch political local… technology organiza- ence. OTA made sure that philosophy and the ability to tion works most effectively boundaries were crossed and work collaboratively. He fo- when it percolates bottom up the direction to congress cused on the technology, but from the plans of people who were always here is what the not the practices that would will pay for and use it.” technology does. Here are allow it to foster innovation. the implications and here is Many believe this is why Rus- But there is also thoughtful who “wins” and who “looses” sia never experienced an en- counsel, oversight and ac- if this legislation passes. In lightenment. countability at every stage, other words informed guid- ending with the committee of ance in the public interest - And that is the lesson and the “Wise Old Men” at the top. If laying out all the points of question for the United you look at the chart on view. OTA was in existence States. pages 21 they are the final for almost 30 years and was step in advising before in- a model for many European The United States is at a vestment is made. And af- approaches., turning point, it’s progress terwards they are also the faltering, its debt growing, its ones who examine if the No more. Its demise opened infrastructure beginning to promised results have been the door for the incumbents crumble. Recall that in 2009 delivered. to use their huge cash flow in the American Society of Civil their very narrow economic Engineers rated America’s The US Congress interest. With no John Ken- Infrastructure and gave it a Ofﬁce of Technology nedy asking “not what your “D”. country can do for you but Assessment what you can do for your country’ and only worship of

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 97 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 private good and free market conduit for internet to the a community network and for the past 30 years. With masses during the 1990s regulatory point of view. In- the cold war “won” we have during a period marked glob- volved are Narda Jones, Chief again lost the ability to think ally by deregulation and pri- Strategic Analysis and Nego- large. A national agenda vate good free market mania tiations Division, Interna- comparable to sending men - the emergence internet tional Bureau and Irene Wu to the moon is not attainable technology, although poten- Yahoo! Fellow in Residence in the midst of partisan ran- tially a general purpose tech- Director of Research, SAND- cor. nology, became locked within MNIA International Bureau a mesh that is seen -- in the Federal Communications Examining the Public USA and to a lesser extent in Commission. the rest of the world -- as the Good versus Private realm of private good finan- I have personally talked with Opportunity cial capital and therefore some of the National Broad- something in which govern- band Plan team. I have been While public versus private ments should not be in- extremely encouraged by the good is a useful metaphor to volved. experience. The political think of in terms of defining process didn’t get all of them telecommunications access, Because of the predominance on board until roughly half because telecommunication of private good thinking we the assigned year was gone. access costs something to keep more to our narrow si- Another example of the situa- provide, it may not fit very los and are very much less tion where the U.S. has built snugly into the category of equipped to take advantage up a destructive momentum, narrowly defined public good. of the thinking on which the which is not proving so easy See: Netherlands builds. to reverse. http://en.wikipedia.org/wiki/ Public_good And therefore extending What has to change? Googin’s paradox to the Nevertheless, roads, water- whole world -- private good The ICT ecosystem and cul- works and electricity also are thinking keeps you not only ture of the Netherlands is, of not free and society does from sharing the benefit of course, reflected in many provide those on a subsidized optical technologies…. it also technology communities in basis and one not motivated locks you into not invented the U.S. We do not have a by pure profit. The logical here way of looking at reality great technology commons and absolutely necessary an- and handicaps your ability to here, but we have literally swer to the “fit” is infrastruc- collaborate. thousands of smaller com- ture and in the USA the 1934 mons blooming everywhere act recognized voice phone in the United States. service as infrastructure and Changes at the FCC acted accordingly. The jury is still out at the To create a successful national infrastructure for However on the way to FCC. But there is reason to America for the 21st Century Googin’s paradox, the picture be hopeful. There is an inter- will require a fundamental re- of what was and what was national group at the FCC as- examination of American val- not infrastructure became signed to work on a compari- ues and governmental. So muddied. Because the tele- son of twenty foreign coun- far, it appears this is a very phone companies became the tries’ broadband policies from

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 98 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 large obstacle. But this does trivial. It will be widely dis- lectual and subject matter not mean that all of Nether- cussed and often influential. strengths of the Netherlands. land’s innovations will be ig- Many will chose to join these It is designed to create a ge- nored. new collaborations. As a re- neric commons overlaid on a sult they will be able to take network fabric with the basic The E-Science Suite Is advantage of a more sophis- tools that individual disci- Not Just for Science ticated, more shareable infra- plines need to ply their trade. structure. This will allow It is being done likely for the them to access, share and first time through a commu- Initiatives like GLIF are, of work with enormous data nications web of planning course, already been used for streams and structures. that gathers the stake hold- international collaboration for ers together with the intent doing science. The scientific We do not believe the e- of tuning them into collabora- community, will as it has al- Science Suite tools and proc- tors rather than competitors. ways done, continue to have esses will be used only for a more open-mind than most science. It is well suited, as of American Enterprise. But Is an ICT research Infrastruc- development continues, for ture for a nation tantamount what will be adopted, the many 21st Century style ac- COOK Report believes, is the to the creation of a knowl- tivities in which competitive new Netherland’s impressive edge framework for a society advantage and success de- new generation of resources, to experiment in new forms pend on a deepening and tools and process for e- of collaboration, problems more seamless collaboration. solving invention, creativity. Science. Internally, we’ve been calling it the e-Science The Netherlands has built a Nobody is calling VL-e’s e- Suite. It’s our name, not one multi dimensional matrix for Science Suit the 5th Para- the team in the Netherlands the application at one level of digm. But over time, we uses. We fell into using it be- Renan’s law of connecting wouldn’t rule it out. everything and Baken’s cause their approach was to, step-by-step, create a grow- metaphysics of holon’s. ing suite of generic middle- The Netherlands vs Building as infrastructure ware, shared tools, support the USA in an educa- means that the Netherlands for collaborative solution organizes what it does in new tion that is part of the ways and does so it more ef- solving, and various modali- larger knowledge ficiently. Approaching it as ties to support a more multi- infrastructure helps pry par- disciplinary approach. What infrastructure ticipants out of disciplinary Bob Hertzberger and the VL-e An example of the disparity is silos. team have done in Amster- ICT support for education in dam is what we call in Amer- the context of overall knowl- ica “heavy lifting”. In the USA what we have is edge infrastructure. too often pork and plunder when it comes to ICT support The problems the e-Science The Netherlands have de- for education. (The following Suite solves, in terms of signed something like a Linux url takes you to the archtypi- making it easier and more OS for all aspects of the cal story of IBM supplying affordable for new domains of Dutch education, research, Internet for El Paso Schools, science and communities to technology and product de- rigging bids, building too begin work with and share velopment. It is a mecha- complex a system, charging enormous databases, are not nism designed to fit the intel- too much, and after the bank

© 2010 COOK NETWORK CONSULTANTS 431 GREENWAY AVE. EWING, NJ 08618-2711 USA PAGE 99 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY - MARCH 2010 was broken, pulling out the Moreover to ensure 100% bent must accept structural infrastructure and leaving the return on investment for the separation or have it imposed schools without connectivity.) incumbents when ordinary by the government in the www.nytimes.com/2004/06/1 savings accounts may pay name of a sustainable future 7/politics/17computer.html?p one half of one per cent is for society. Just as schools agewanted=all not the idea of equal justice were ordered desegregated in for all on which the Unites 1954. Unfortunately this is only one States was founded. Secon- of the problems blocking pro- dary schools will also remain And the pace of change ingress. Read on. locked out of the benefits creases. It never slows. Just that have been described in when things look darkest The power of this paper. Why because the come signs of the dawn. In Research Networks in FCC created the school and 2005 David Isenberg held the libraries fund in such a way first of five conferences on the Netherlands. that it operates for the bene- the political issues around fit of the incumbents scarcity connectivity and access to It has lead to a not fully un- model. They are forced to resources in Washington DC. derstood dichotomy where become customers for the The theme was Freedom 2 research networks are trans- incumbents and prevented Connect. forming what they touch in a from using universal services way that the current public funds to establish their own Five years later, on Thursday stasis renders impossible for service at less cost. They January 21 2010 Hillary Clin- the rest of us to grasp. Not were forbidden to offer their ton enshrined these three connected to them we cannot own services setting up a two powerful words into official experience them first hand. billion dollar a year feeding US policy. We in effect don’t know what trough for the incumbents. we are missing. In the past, The consequences were pre- the catch-word has been pa- The Freedom to Con- cisely what Dave Hughes the tience, it will come to you. nect “cursor cowboy’” predicted Well this time the truth is it when the program was estab- won’t. The incumbent busi- “The final freedom I want to lished more than a dozen ness model forbids it. address today flows from the years ago. The El Paso story four I’ve already mentioned: above and thousands more The question is how to break the freedom to connect – the were inevitable. out? idea that governments should not prevent people from con- If R and E networks succeed Because if we don’t, we are necting to the internet, to in bringing this future to deprived of not only of the websites, or to each other. some campuses but not to all SURF sponsored technology The freedom to connect is education, I contend it is an transfer described earlier in like the freedom of assembly invitation to another separate this paper : in cyber space. It allows indi- and unequal case analogous http://www.jisc.ac.uk/whatw viduals to get online, come in the United States to the edo/campaigns/res3.aspx together, and hopefully coop- Supreme Court 1954 decision and The Future of higher erate in the name of pro- abolishing school segrega- Education: gress…. tion. Don’t misunderstand http://bit.ly/5ejtEl. me. The R and E networks must succeed and the incum-

The spread of information “What we must help teach computer effectively van- networks is forming a new them is how to think and how ishes. Computing in effect nervous system for our to collaborate via cell phones, will merge with the physical planet. When something tablets and laptops. Strong world. Newspapers for ex- happens in Haiti or Hunan, community network links and ample will show live informa- the rest of us learn about it in 'weak links' will appear be- tion you will be able to pick real time – from real people. tween them and stabilize the up a book in a book store and And we can respond in real world. No dictatorial govern- your interface will show you time as well… ment, bureaucrat, spin doc- and tell you whatever you tor, priest or mullah will be would like to know about the As we sit here, any of you – able to stop the kids doing book. In effect “these devices or maybe more likely, any of that.” by making the machines dis- our children – can take out appear wind up helping us to the tools that many carry Having gotten “freedom to be more human so that we every day and transmit this connect” accepted as part of do not wind up being ma- discussion to billions across the bed rock foundation of chines sitting in front of other the world. official US policy is a very machines.” important achievement on http://www.ted.com/talks/vie Let us make these technolo- the part of David Isenberg. w/id/685 gies a force for real progress But the operative word is the world over. And let us go connect and that leads to Re- Pranav Mistry leads a group forward together to champion nan's law. Everything wants at the media lab and India, these freedoms.” to be connected. Everything that is developing this tech- http://blog.austinheap.com/t works better when connected nology. The hardware is ranscript-of-sec-clintons-spee and that the absence of con- commodity. The code is ch-on-internet-freedom/ nection is death. open source. Because they cannot be locked in silos, the Two days later writing from But We End with Hope walls of scarcity we have the Netherlands Jaap van Till been discussing in this book elaborated on what this Renan’s law however is be- will either dissolve or become means. coming more and more evi- irrelevant. Renan’s law will dent. Augmented reality and triumph because things in- “The young of this world, something called “the sixth deed work better when con- emancipated by online net- sense” are rapidly emerging nected. The Netherlands has working, do no longer want technologies. As a result our built the receptive environ- old men to tell them what to movements in the physical ment for what Renan and think.” world will blend with the digi- Mistry describe. The United tal world with much less ef- States must catch up. fort and very quickly the

Contents

I. How the Dutch Did It Right p. 1 When so many others get it wrong

II. The Netherlands National ICT Research Infrastructure p. 5 History, character, policy and pragmatism

III. The Direction of ICT Infrastructure in the Netherlands in p. 24 Early 2010 An interview with Wim Liebrand, Kees Neggers and Cees de Laat

IV. The Ascent of e-Science p. 41 Promise of the 4th paradigm

V. Making E-science Work: The Middleware Solution p. 44 An interview with Bob Hertzberger

VI. Growing E-Science Domains for The Netherlands p. 64 Roadmap for a next generation of science

VII. Potential Customers with Global Agendas p. 68 An interview with David Zakim, MD

VIII. How A Progressive ICT Infrastructure Beneﬁts the Economy p. 78 The innovation engine as open fabric

IX. SURF as an Economic “Midwife” for Technology Transfer p. 85 An interview with Hans Dijkman, Kees Neggers and Bob Hertzberger

X. Coming to Conclusions p. 91 Realizing the beneﬁts of a re-usable infrastructure

XI. Re-thinking American Infrastructure p. 95 What has to change

Cook report on Internet Protocol February 2009

THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009

ICT and E-Science as an Innovation Platform in The Netherlands A National Research and Innovation Network What Can the US Learn from Dutch Experience?

“The dogmas of the quiet past possible are carried out by Volume XVII, No. 11 are inadequate to the stormy decentralized groups. February 2009 present…As our case is new, so ISSN 1071 - 6327 we must think anew and act What can be done privately amine in detail their current anew.” Abraham Lincoln should be. But government research and innovation net- must exert oversight and in- work infrastructure. By means of an examination sist on transparency. We can of research networks in Hol- only hope that the new Ad- In SURFnet6 the Dutch have land, this issue presents ministration will begin to ex- likely the most innovative some ideas for ways in which plore these and many other optical NREN in the world. To an American National Re- new ideas. leverage the capabilities of search, Education and Inno- SURFnet6, in 2005 they vation Network could be Before Resource Use started ICTRegie – the Neth- structured. Careful Deliberation erlands ICT Research and In- novation Authority. The mis- For the first time in more and Coordinated sion of the Netherlands ICT than a generation, the model Planning Needed Research and Innovation of unregulated speculative Authority www.ictregie.nl is financial capital has shown its In the area of networks as an to enhance the innovative bankruptcy, the ability of integral part of national so- strength of the Netherlands government to encourage cial, economic, and research the coordinated use of soci- infrastructure The Nether- ety’s resources in the public lands is now a world leader. interest should become a fo- The Dutch are building a na- On the Inside cal point of our political life. tional, largely open, fiber infrastructure. As we demon- Pioneering LightPath By encourage I don’t mean strated in our January 2009 Networks for Global dictate, but rather to try to issue, The Netherlands has a Science Collaboration Contents p. 65 act on behalf of agreed upon pragmatic way of finding out basic principles that wherever “what works” and then just Please read Explanatory Note doing it. In this issue we ex page 64

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 1 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 by means of information and throwing huge amounts of promising innovations using communications technology. money in an uncoordinated ICT, encourages them to Two key objectives derive way and at a wide array of seize the opportunities for from the mission statement. problems in the hope for innovation, and brings those First to introduce unity and some results. The modern opportunities together within consistency to the strategic world is too complex and a national vision. direction of the ICT research capital in too short supply for and innovation by means of the continuation of this prof- 4. Making the best possible the development of a national ligate way of governing. use of self-organization abili- strategy that enjoys broad ties of the field, ICTRegie de- support and, second, to en- Here is a world view we need velops new instruments and sure ongoing strengthening to emulate: programs in selected fields. and appropriate dynamism of a Dutch ICT knowledge infra- The four pillars of ICTRegie’s structure geared towards strategy E-Science high social and economic 1. The Netherlands ICT Re- yield. ICTRegie has a Strate- search and Innovation The remainder of this issue gic Plan 2005-2010 found at Authority (ICTRegie) was focuses on E-science as de- http://www.ictregie.nl/index. founded with a view to veloped under the leadership php?pageId=6&l=en&person strengthening the country’s of Kees Neggers and Cees de eelId=&pubId=15 ICT knowledge infrastructure Laat at SURFnet in conjunc- (in both focus and mass) and tion with others in the United This Plan is designed to es- to match supply to demand in States (including Tom De- tablish a framework for think- terms of knowledge and ap- Fanti, Larry Smarr, Maxine ing, discussion and coordi- plications. The overall aim is Browne, Joel Mambretti),Ja- nated action to ensure effec- to enhance the innovative pan, Korea, China and tive use of economic capital ability of the Netherlands Europe. invested in the goals of the (impact). organization. It strikes me as Kees Neggers sent me the somewhat similar to the 2. Within the complex field of December 11, 2008 ICTRegie Japanese ICT plan of 2000. ICT, its many research disci- report called “Towards a Other plans from other coun- plines on the supply side, Competitive ICT Infrastruc- tries could be identified. countless application domains ture for Scientific Research in What troubles me is that this on the demand side and a the Netherlands”. This report approach in the United States large number of organiza- can be found at: would be likely to be criti- tions throughout, a purely http://www.ictregie.nl/public cized as state planning. Of top-down style of direction is aties/nl_08-NROI-258_Advies course the five year plan of unlikely to be effective. _ICT_infrastructuur_vdef.pdf the 1930s remains a joke. 3. Three accordingly, ICTRe- The report evaluates and What we don’t seem to have gie mobilizes stakeholders on builds on both SURFnet’s and considered and I suggest that both the supply and demand the e-Science Virtual Labora- we should consider is side, encouraging them to tory project’s innovation ap- whether a plan for coordi- join each other in thinking proach. I refer readers to nated use of resources in a about opportunities for inno- the conclusions in sections framework for such can be vation. It challenges them to 6.8, 6.9 and 7.0 found on far more productive than inspire each other to arrive at pages 59 and 60 of this is-

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 2 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 sue. Trying to place my storage, grids for real time from theory to practical emerging understanding of e- collaboration of people. and product, if must depend on science in context I asked instruments designing pro- these e-science platforms? Kees by e-mail jects around the world - it seems to be the internet Neggers: Very likely - yes. “Let me test my understand- based emerging global plat- We see a growing demand ing: There is form for all science, teaching emerging in all disciplines. and investigation globally? The Dutch Roadmap Commit- 1. theoretical science tee for ESFRI already put 2. applied science (building Neggers: I agree, actually it more emphasis on alpha and something from theory). is a push to a "better inter- gamma science and as a re- 3. computational science net" too. sult on data versus computa- building a computer model to tion. This will not replace the see if the application works The ICTRegie report defines need for computation of and finally e-science in chapter 6.5 as "a course. It shows that infra- 4. e-science development to bridge the- structure like networks, com- gap between scientist in ap- putation and storage, plus E-science is not really an in- plication domains and the the required software and dependent 4th branch of sci- development of ICT." In pre- know how will be essential ence. In reality it is more senting the report, the direc- tools for researchers in all like a platform or foundation tor of ICTRegie called it disciplines soon. with which to enable the ef- enhanced-science. In the forts of scientists at all three preparation of the report And that’s why it should be levels to support the eco- some suggested to avoid the made available, at high qual- nomic foundations of their word e-science completely ity, in its own right. To avoid society. because of its unclear mean- duplication and/or delay in ing. realizing it. How to describe it in more detail? Using light path net- COOK Report: But if theo- works - OptiPortals - data retical science ever advances

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 3 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 Adopting a New National Infrastructure How Fiber Became the Basic Platform for Education, Commerce, Technology Transfer, and Research in the Netherlands

Editor’s Note: I would ar- cations law. In the new law, role to streamline the dig- gue that The Netherlands has which came in effect on De- gings. Another important as- a more complete and perva- cember 15, 1998, all land pect of the new telecom law sive fiber optic infrastructure owners, both public and pri- is that when you make use of than any country in the vate, had to tolerate the dig- your right of way, you have world. On the citizen side ging of fiber by all public to allow sharing of the dig- fiber to the home and apart- network operators. Public op- ging. In other words if you ment and on the research erators did not have to pay begin a project you must an- education and enterprise side for use of right of way. Until nounce what you plan to do not only fiber but also light than, only the incumbent op- and anyone who also want to path networks to 160 univer- erator, KPN, had these rights. lay fiber as part of that exca- sities, research institutions Also operators of non-public vation is entitled to do so as and enterprises. This issue networks were allowed to lay long as that entity shares the revisits SURFnet for the first there own fiber, but they had cost of the excavation appro- time since the March-April to make an agreement for priately. issue of 2005. this with the land owners first. City Governments were This all meant that many In this opening interview, given an official coordinating more were able to play and, done on Monday No- vember 17th at Super- computing 08 in Austin Texas with SURFnet Director Kees Neggers, we look at the events that enabled the Neth- erlands to become both the optical and IP focal point for Europe.

Sparking the Fiber Revolution

COOK Report: How did the fiberization of the Netherlands begin?

Neggers: In the 90s the Netherlands liber- alized its telecommuni-

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 4 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 especially in the end of the in the Netherlands were very Cisco GSRs, we wanted our 90s, a lot of fiber was laid. well interconnected. customers to be able to con- Particularly in Amsterdam. De Laat: The first two loca- nect at 1 Gb or 10 Gb speeds See picture from an old pres- tions for the Amsterdam as well. We had asked the entation at that time above. Internet exchange were ac- providers for Gigabit Ethernet All of this was put in before tually both within the Science connections, but operators the breaking of the .com Park. were unable to provide these bubble. All this excavation in 2001. Of course, to deliver included a large number of Neggers: We started in two the gigabit Ethernet service, empty ducts that could easily places - at SARA, that ex- the operator would have to receive additional fiber. change was run by us and at bring in fiber. Realizing this NIKHEF, a Dutch High Energy we decided why not just ask COOK Report: If there are Physics site. We designed an for the fiber! We then asked disputes, what happens? exchange architecture with a the operators to give us the high priority on resiliency - fiber pairs we needed for our Neggers: The cities have to with a policy that encouraged locations. Again, they refused deal with disputes for fiber people to connect in both lo- to deliver. digging activities in all public cations. After a few years we grounds. OPTA, the Dutch formed an association to cre- We also realized that individ- telecom regulator, was ate a level playing field for all ual trenches dug to single charged to deal with disputes participants. In order to be locations for every SURFnet5 for digging in private territo- connected to the exchange member would not make ries. In both cases of course you simply had to join the economic sense. Therefore also the standard legal ap- association as a member and we planned fiber rings peal procedures are avail- abide by its rules. AMS-IX is through the areas where we able. still a not-for-profit asso- had clusters of institutions ciation and is not con- connected to SURFnet5. We The Signiﬁcance of trolled by any single did this in conjunction with AMS-IX member. city governments which joined in the procurement so Because a neutral exchange that they could also connect Neggers: The Amsterdam existed in Amsterdam very their institutions to the fiber Internet Exchange (AMS-IX) early on, as the Internet bering. began in the early 90s when came more important, it the Internet was primarily a served as a magnet that en- The operators did not research network. But early ticed operators to bring in even respond to our ten- on commercial players were connections. To take advan- der which of course left us allowed to interconnect and a tage of the conductivity, all no choice but to go to the very important decision was they needed to do was to specialized fiber installa- that the commercial carriers bring their fiber duct into the tion engineering groups to were allowed to interconnect AMS-IX location which then which the operators had with the early research net- served as a catalyst for this outsourced their own works on a neutral basis. This continued growth. building before the crash meant that, very early in the caused them to stop fur- game the research networks And when we had SURFnet5 ther capital investment. and the commercial networks up and running in 2001 with 10 Gb Lambda’s connected to

Due to the dot com crash, fiber laying work would tions in dark fiber and only the operators were no longer probably not have been needed some additional dark investing in capital expendi- available to us. And another fiber in the backbone. This tures and consequently, the nice thing was that these was the difference between companies to which they had companies were the ones us and other NRENs that are outsourced the building of who originally had done the going hybrid now. These their own fiber networks were digging for KPN, AT&T, Level NRENs all start with the standing idle and were in 3 and other carriers so they backbone and then still suffer need of work which we were knew very much what they from the last mile problem. glad to give them. Because were doing and were quite of these economic conditions capable of providing good Because we have an all opti- the first ring with a length of work for us. cal network, we were able to 40km with 20 locations and design it so that we had opti- 96 fibers did cost only €1 mil- It was also fortunate for us cal nodes at our largest insti- lion. The return on invest- that the operators were no tutions. Consequently, if we ment for the entire ring was longer like the old PTT be- needed to bring in additional less than four years based on cause most likely they would connections, it was easily do- 2 Mb per second leased line then still have completely able and all-optical as well. prices for these 20 locations. owned and controlled all en- The cost of SURFnet6 was gineering installation opera- similar to the cost of SURF- Consequently this was a notions. net5. SURFnet5 was a 10G brainer. The carriers soon backbone with 15 big GSR discovered that they would COOK Report: Gradually routers. With the all-optical have competing fiber rings then, by means of this strat- approach we reduced the over every square inch of egy, you had all the fiber number of router backbone their territory if they did not procured that you needed? sites from 15 to two. We make their own fiber more used two for resilience. Both available, which they then Neggers: Yes, and by about function all the time and eve- did. This was a tipping point 2005 we had connected all ryone is connected to both. that actually opened a dark 160 SURFnet member institu- We also have a border router fiber market for everyone in tions directly by fiber. This at each of our two backbone the Netherlands. was local loop dark fiber with sites to separate external a 15 year IRU owned directly from internal routed traffic. Now, had they provided by us. We completely solved So the whole SURFnet6 IP those gigabit connections, the last mile problem by network is based on 4 routers we might have become means of these local loops. in two sites. hooked on them and not have started to build them But then, when the contract COOK Report: It was this for our own use. We with British Telecom ended idea that was the germination never thought that having and the SURFnet5 backbone, of your hybrid network ap- to dig ourselves would be- based on BT 10G lamdas, proach? come so attractive. Of had to be replaced, we said; course, had it not been for why not put a little bit of ex- The Germination of the “dot com” crash and tra fiber in the backbone and the Hybrid Network things were booming, then go all optical? This is what the technical civil engineering became SURFnet6. In effect Approach in 2001 companies who are doing the we had all last mile connec-

Neggers: No, the seed for location where our astrono- country. We installed their hybrid networks was in 2001 mers are and where the Avici routers but unfortu- when we had these 10 Gb European JIVE correlator is. nately they withdrew from lambdas from British Telecom (You see that some time ago the market. Nevertheless and realized we could do a lot the Netherlands was rather with our collapsed backbone more with them, specifically interested in navigation and design we have only two such for demanding users like therefore we invented all routers connecting all our high-energy physics and as- sorts of tools like telescopes customers and we were able tronomy. We also realized and still astronomy remains a to make a transition to new early on that the research significant research field in Juniper routers that was world is an international the Netherlands.) transparent to our users. We community for which we had installed them back to back to develop lambda network- Global Crossing, which was with Avici’s and every week ing on an international scale also having tough times in moved a number of 10 Gb from day one to avoid inter- 2001, had a cable from Am- connections from one plat- working problems in the fu- sterdam to Hamburg that form to another. We did so ture. Consequently in 2001 went within 40 km of Dwin- in a maintenance window so we first ordered a 2.5 gig geloo. Global Crossing was that customers did not notice from Amsterdam to Star- prepared to take a fiber pair the change. Light, the US research net- from Amsterdam to Hamburg working exchange point in and open it up at the point In the early 2000s, when we Chicago, in order to be able nearest to Dwingeloo, if we’d were experimenting with light to explore these options with dig a 40 km trench from that paths from the Netherlands Tom DeFanti and Joe Mam- location to the Dwingeloo Re- to Chicago, there was still a bretti. search Center. lot of hesitation in Europe. I remember that I had a pres- By iGrid2002, which was in Adoption of entation at the NORDUnet Amsterdam in September Lightpaths - Not only conference in April 2002 ti- 2002 we already had two 10 tled “Research networking – Gb lambdas between Am- for Science but also The next Phase” where I said sterdam and the US as well for ICT Departments this is the future this is where as the 2.5 gig wave. But no- we have to go. Even NOR- COOK Report: So where body could really use them at DUnet was not ready for this have you taken your SURFnet that time. All of the TCP/IP this at that time. Most people architecture in the last three stacks and the workstation thought that 10 Gb IP, which years? Ethernet cards crumbled. was already running in They had all been testing the SURFnet5, would be the fu- Neggers; The SURFnet6 ar- stacks in the lab where they ture and bring plenty enough chitecture worked out pretty worked but suddenly with for everyone. But we went much as we intended. Nortel these long distances they on this path anyway because won the procurement as you failed. However, everyone we were convinced that we know. We were the first cus- realized the potential and the had no choice. Our astrono- tomer for their new Common lambdas themselves were mers were already telling us Photonic Layer (CPL) equip- there to stay. that this was their only fu- ment and we are still very ture: by having telescopes happy with that. The network In 2001 we also planned a connected at high bandwidth is totally optical all over the dark fiber to the Dwingeloo to the JIVE correlator they

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 7 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 could do real time correla- COOK Report: This is very the big universities and a tions. So we also connected interesting because it must few large research institu- our 16 linear array antenna mean that there are potential tions would be connected radiotelescope in Westerbork lightpath applications in con- to the lightpath network. to JIVE in Dwingeloo. necting remote corporate Now however we already campuses with each other via have about one third of Consequently, when we optical private networks? our connected institutions started developing ideas for using lightpaths. and SURFnet6 in 2003, we had Neggers; Absolutely. If you most of these institutions this footprint of all-fiber- can have a centralized ICT are using lightpaths to connected institutions and it department concentrating create Optical Private was not a to big expense for your services at one or two Networks for e-learning us to do an all optical hybrid (for resilience) places then, and administrative serv- network to provide both opti- as a result, you can have the ices. On a domestic basis cal and IP services. For us it same quality of service to all the uptake is ICT depart- was a no-brainer. But we re- your researchers or, as the ments. While on an inter- alize that for everyone else case may be, to all your re- national basis the uptake who has not yet built the fi- mote corporate campuses. is all researchers. ber infrastructure to do this it Now the people producing will still be a major invest- software for these services to Internationally we have GLIF, ment. But today there is no remote campuses will find the Global Lambda Integrated escape for anyone involved in that they need to take care of Facility, collaboration where research networking, and if the latency problems. we share lambdas and coor- you do not do this, you may dinate the scheduling of just as well walk away. For example with iGrid2002 lightpaths for experiments. we found out that they had This is very successful. Right When we built this hybrid not thought about the prob- now many researchers are network motivated by our big lems of remote users scat- making serious use of science users, we found to tered over a very large wide- lambdas internationally our surprise that the biggest area network. With ordinary where IP routed traffic at uptake in the Netherlands speeds and TCP a dropped such high speeds and quality was not with the researchers packet didn’t matter. But is really not feasible. Pay- but rather with the ICT de- with gigabit or multiple giga- ment problems for the re- partments. The reason was it bits light waves running on searchers have been avoided is the ICT departments that TCP if the packet was so far because we have had a are responsible for providing dropped the stack would lot of donated or shared network services through to crash. We also discovered lambdas for international ex- remote campuses for e- that login protocols for access periments in use. learning. They had problems to remote servers not always doing this by means of a know how to deal with WAN But nationally, even in the routed IP network. They saw latencies. Netherlands, there is still a that they could order light- lot of help needed to make paths from us, and use them In the Netherlands, when the facilities available to and to extend their local area we started SURFnet6 we used by the applications peo- network, and their problem estimated that by the end ple, that is corporate scien- was solved. of the GigaPort project in tists or university research- December 2008 most of ers. Researchers are faced

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 8 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 with local loop problems in- connected institutions. But gin to function as an ap- side the campuses, needs for we are not in the business of plication broker because reallocation of ICT budgets, doing this commercially be- the network is already and last but not least, re- cause we believe that there there. To do the brokering structuring of their research are enough other people in all you have to do is make projects to be able to take the Netherlands who can do sure the parties, or their advantage of the new facili- it certainly as good as we can providers, are connected ties. All this takes time and and perhaps better. In other to NetherLight. effort. We therefore have to words doing this is not a part do much more to stimulate of our core business. COOK Report: To what ex- the uptake of the optical tent and do you have good network by the research Dutch to American cooperation between univer- community. sity and private industry Analogies and the E- where the desired end is Nevertheless the fact that Science Virtual technology transfer? about one third of our cus- Laboratory tomer base in two years time Neggers: Cees de Laat can is actually using lightpaths, When you consider what you now take over because he is we rate as quite a success might like to try in the United part of the Dutch Virtual States you have to realize Laboratory for E-science (VL- COOK Report: Are you in a that the situation in the e) project, which is the other position where you could Netherlands is very much dif- project that, in conjunction provide training to people ferent than in the US. NL is with GigaPort Next Genera- outside of the Netherlands very much smaller. In the tion Network (the project that who wanted to embrace this US terms the Netherlands is developed SURFnet6’s optical technology? very much like a regional US network), was funded out of network. But remember the government supported so Neggers: As you have seen that in addition to the con- called Bsik program. When we, and the research net- cept of the Netherlands as a we started the GigaPort pro- working community in gen- regional there is also one ject in 2003 we joined forces eral, are quite open about more huge difference. All our with the University of Am- what we do and how. This is research institutions are con- sterdam’s VL-e project. We also quite essential. Other- nected by fiber. There is no submitted a largely comple- wise it would not be possible last mile problem towards the mentary proposal and said, if to create a transparent multi- institute anymore. we are both funded Cees his domain lightpath service to group’s networking activities our international user base. If anyone wants to connect to will be in the GigaPort pro- a Dutch university or re- ject. But he also had some GLIF of course is an excellent search Institute, the only non-networking responsibili- vehicle for this and I would thing they have to do is bring ties in the virtual laboratory recommend that anyone in- a lightpath into NetherLight, e-science project and this is terested should join this the Amsterdam Glif Open the area which your question community. Also users are Lightpath Exchange. Conse- now directly touches upon. welcome to join the GLIF Re- quently in the Netherlands search and Application Work- a broker with knowledge Cees de Laat: The Virtual group. SURFnet is organizing of industry and the re- Laboratory for E-science (VL- seminars and training for our search area could just be- e) has university partners

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 9 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 and industry partners. On the scribe the intentions and re- lem. This is a focal point that industry side for example sources that the commercial by means of our grid infra- there was Unilever, IBM, and companies and universities structure and parallel proc- Phillips, the Dutch electronics bring with them when they essing software we can help company that operates a sit down at the table to talk them successfully tackle. large scientific research park about working together. Do They need to share data with in Eindhoven. Now the BIG- they just look for the ability each other and to do data GRID project, a sister project of their staff scientists to mining of the data they cre- of VL-e, has build a consider- have general day to day in- ate. You need Web 2.0 tech- able grid. Part of this infra- tellectual relationships with nology and metadata capa- structure is installed in Phil- their university counterparts? bilities to describe the raw lips’s Research park in Eind- Or does it get more specific data and make it more ac- hoven. and do they say we’re looking cessible. This also needs for a new kind of algorithm to ontologies and, as a result, The industry uses this infra- do some specific task what one of the group’s at VL-e is structure to enhance their talents can you bring to bear working on needed ontologies own methods of handling it and perhaps contract-based to describe tastes in a food their own scientific data. research to help us achieve database from Unilever. If And even Unilever which is a this end? you want to do data mining big food and chemistry con- and cross correlation be- cern in the Netherlands uses De Laat: They develop work tween groups of people with it to handle their food infor- packages and within those tastes, this infrastructure is matics data. They use our E- work packages there are really necessary to obtain science technology to go tasks for the different part- that goal. through their food informa- ners. As a result they can tion databases to answer build a very formal project. COOK Report: I think I am questions like: in what ways hearing from you that the do which people taste what But what industry really truth of the matter goes far they eat? Phillips uses it in finds interesting is that beyond the assumption that, their medical department we bring to them a basic with the big pipe connecting where they have made medi- infrastructure based on a corporations with researchers cal implementations and also grid computer architec- and universities with re- use remote experimentation ture enabled by high- searchers, the scientists on technology to enhance their speed optical networks both ends would figure out products. As a result we and containing a layer of some way by which they have close working rela- very useful middleware. could collaborate with each tionships with several In the middleware layer you other. large companies many of have software controlling the which are also working networks but the same of De Laat: We do things dif- together with SURFnet in course holds true for the grid ferently. Here is a slide [see university-based private middleware layer that in- bottom of next page] that sector partnerships. cludes Globus for applications describes the infrastructure scheduling and distribution. that we have built to enable COOK Report: You mention serious ongoing E-science E-science and databases and One of the things that indus- research cooperation be- the optical network and I try is typically very interested tween universities and enter- wonder how you would de- in is the data explosion prob- prises. This is the slide from

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 10 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 the slide deck called Beyond to be used by the users. COOK Report: This is very H y b r i d N e t w o r k i n g There is software needed as interesting because the ap- http://staff.science.uva.nl/~d a layer between the network proach that I wrote about last elaat/talks/cdl-2008-11-14-F. resources and the users to summer was much more nar- pdf and I’m showing the enable the use of the entire rowly focused in the belief fourth from the last slide that package. The users are in that a company would iden- is about e-science. different disciplines repre- tify a specific problem to sented by these smaller ver- solve. It then needed to go At the foundation layer you tical bars and what is becom- to a specific researcher and have a grid substrate which is ing much more widely ac- develop a narrowly focused the physical network. Above knowledged now is that the contract applicable to the that you have the middleware most attractive areas of re- task and that, somehow in high-performance distributed search are not within these advance, everyone on both computing web and grid. disciplinary silos but rather sides of the fence would within the interdisciplinary know what was needed. Neggers: Supercomputers, spaces between them. telescopes, microscopes and The Optical Network whatever else you need as De Laat: It is the combina- and its Platform for the basic tools required for tion of these disciplines that your research are considered gives new science. Collaboration Become to be in the chocolate cov- Basic Infrastructure ered box called high- Neggers: And it is in these performance computing and little dark areas between the Neggers: But more impor- storage. All these resources disciplinary silos that the new tant for the Netherlands, in the chocolate brown bar inventions take place. and switching back the are connected to the network focus to us, is that the Gi- gaPort project is now declared a large success by international reviewer’s and by our own users. Our government has been informed by the reviewer’s and especially by the Dutch Roadmap Com- mittee for Large Scale Re- search Facilities, which advises the Dutch Gov- ernment on our participa- tion in the European wide ESFRI projects that this network is needed in the future regardless.

They were told that an advanced network is necessary no matter what you wish to do and they have

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 11 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 been given the advice now whether they are in universi- Neggers: In our view in- that SURFnet should no ties or corporations, will be novation needs the net- longer be funded out of able to rely on the high- work and hardware re- project funding but rather quality ICT infrastructure sources plus the software basic infrastructural fund- needed that includes a hybrid and the people that have ing is needed to keep network, high performance the knowledge of how all SURFnet in place. computing and all the storage of these layers interact you would need as a founda- with each other and in ad- It seems very significant that tion for their further work. dition to this knowledge the government has recog- you still need a major out- nized this and the Ministry of COOK Report: How much reach effort to involve the Science and Education has understanding of what you users for which it has now prepared structural have just described to me been created. funding on an ongoing basis exists here in the exhibition for SURFnet’s continued role hall at the Supercomputing De Laat: From the light in scientific and technology 2008 meeting that is the blue box you need people innovation in the Nether- premier high-performance working with every disci- lands. But even more sig- computing meeting in the pline to make it happen, nificant is that they are world? because the people in the saying that SURFnet is not disciplines themselves enough and that we also Neggers: Some but proba- cannot do that. You can need software to glue this bly not enough. The key is- not expect the biology together and the ICT com- sue is that governments professor to understand ponents that have enabled need to understand this how to go out and identify SURFnet and the VL-e pro- and not just the practitio- and hire the people neces- ject to succeed need also ners at this meeting. sary to teach him and his to be funded in the future colleagues how to use the on a structural basis and We consider this ICT in- high performance tools. no longer on a temporary frastructure up to and in- project basis, where they cluding the blue layer here COOK Report: In the United allways would have to not just as basic network States if you promote tech- compete with normal re- infrastructure but rather nology transfer from research search projects. as a more general infra- networks to enterprises, and structure for the informa- do it without an infrastruc- Therefore the Minister has tion era. Just as you ture like ICTRegie, you will be asked ICTRegie to make a 10 needed the railroads in dealing with professorial fief- year forecast for the neces- the 19th century, we need doms and the kingdoms of sary integration of the ad- this now and having it will University incubators on what vanced network and the ad- create new jobs and new might be perceived as a vanced ICT infrastructure growth. competitive basis. that enables it to perform its fundamentally important role De Laat: Hardware is very See the article from the New supporting the innovation important for the chocolate York Times found here: necessary to continued eco- brown layer in the middle, http://www.nytimes.com/200 nomic growth and develop- but for the blue layer above 8/09/07/technology/07unbox ment. This means, that in the that, you need people. .html?_r=1&oref=slogin&pag future researchers no matter

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 12 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 ewanted=all Is this a problem COOK Report: What are the Tom DeFanti, Cees de Laat, in the Netherlands? disciplines listed in the small Joe Mambretti, Kees Neggers, vertical cylinders? Bill St. Arnaud: "TransLight: De Laat: Not as much I a global-scale LambdaGrid for think because we are smaller Neggers They are genomic e-science", Communications we have only about 12 uni- markers, biobanking, fun- of the ACM, Volume 46 , Is- versities and a number of damental research into mat- sue 11 (November 2003), technical institutions and we ter, virtual knowledge studio, Pages: 34 - 41 tend not to have universities cognitive science, environ- with vast private endow- mental disciplines earth sci- COOK Report I’m thinking ments. In the Netherlands ences oceanography and the that Obama science transition however we also have devel- final cylinder says “essentially team needs to understand all oped specific groups that are and so on”. This is not meant of this. But of course how to working with the kinds of to be an all embracing list of do it is another issue. computer science that we de- disciplines. veloped to assist disciplines Get the Optical in applying it to their re- Fortunately there is com- Infrastructure Right search priorities. plete understanding within our government Neggers: Get well connected how important this is to Neggers: We do have all this and that point becomes a maintain a modern econ- in place. We do have a hybrid magnet for further connec- omy. network connecting all the tion. From NetherLight, as institutions. We do have a you can see on the map at De Laat: The journal that I high-performance computing the top of page 14, you can gave you is a special issue budget that renews super- go to any place you want to about the application of the computers every four years. in Europe, the US, and Asia. Optiportal to this E- science We do have a large storage We are very well connected infrastructure. grid. We do have E-science as you can see in this section tools for laboratories. All this of the latest GLIF map from is already in place. Larry Smarr, Maxine Brown, May of this year. There is Cees de Laat, Editorial: "Spe- also now a research dark ficial Section: OptIPlanet - The The new thing is that the ber from St. Petersburg into government now prepares OptIPuter Global Collabora- Finland and it means we will to finance all of this on a tory", FGCS, Vol 25, issue 2, soon have 10 Gb Lambdas structural basis and the feb 2009, pages 109-113 from NetherLight to St. Pe- only condition they will tersburg that will then be L. Smarr, T.A. DeFanti, M.D. place on this is that the connected into Moscow. research community as a Brown and C.T.A.M. de Laat, whole in the Netherlands "iGrid 2005: The Global The world is a global commu- will create a national gov- Lambda Integrated Facility", nity now, anyone wants to ernance structure that will editorial, iGrid2005 special peer and work with the very issue, Future Generation make this a well inte- brightest people on the grated effort where all the Computer Systems, volume planet no matter what coun- components will be in bal- 22 issue 8, pp. 849-851 try in which they live. The ance with each other. (2006). networks are the vehicle that we build for users and for

links from NetherLight to MAN LAN in New York too. Any American based innovation effort could take a 10 Gb link from New York to Am- sterdam for about €100,000 a year. The cost per month of a 10G lambda from Am- sterdam to New York is now less than €9000.

COOK Report: But the 10 Gb Lambda for an enterprise would be more than €9000?

Neggers: No, this is a commercial price. It is based on a public procurement. We have had bids from more than 10 different players. So prices between Amsterdam applications to meet world- could buy a 10 Gb link from and New York are very com- wide. MAN LAN in New York City to petitive. But if you want to Netherlight and plug-in to go beyond New York the diffi- COOK Report: While in the- some of the work of your e- culties start. If I want to ex- ory then an American innova- science virtual laboratory? tend the link from Amster- tion network effort focusing What would be necessary to dam to Chicago, it doubles on National Lambda Rail try that? the price of the link from Am- sterdam to New York. Neggers: A very simple answer. First COOK Report: And how do of all, NLR is con- you get from NetherLight nected today via the with a direct connection to many GLIF links con- AMS-IX? nected to NetherLight, including IRNC, IEEAF Neggers: That is very easy. and GLORIAD links. NetherLight is at SARA in the Secondly, at the mo- same building as one of the ment we are doing a nodes of AMS-IX. And many procurement to renew install lightpaths to AMS-IX o u r t r a n s - A t l a n t i c now to take advantage of the lambdas to New York competitive internet up- and to Chicago. NOR- stream prices in Amsterdam. DUnet and NLR are also taking part in this procurement and will Kees Neggers at Sc08 in Austin be getting their own

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 14 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 The Basis for a Future Internet? Optical Hybrid Networks and e-Science as Platforms for Innovation and Tech Transfer

Editor’s Note: I continued slide shows our organization search department of KPN. the discussion begun on No- within the University and the He did a lot of virtualization vember 17th with Kees Neg- topics that my group investi- work on the telecom net- gers and Cees de Laat, on gates on: work; for example you could the 19th with Cees alone. go to a web service, type in My group has four sections two phone numbers, and COOK Report: Where does lead by senior scientists. One have them call each other. your work on optical hybrid of the groups is lead by a Effectively what he was doing networking fit in the context professor from Industry, was making network compo- of the E-Science program at Robert Meijer. nents into software objects. the University of Amsterdam and the Dutch e-Science pro- Rob started out five years He is approaching this by vir- gram Virtual Laboratory (VL- ago with an Internet of tualizing devices and turning e) work we talked about yes- “things” – that is sensor net- them into web services. He terday? works. He came, like myself, encapsulates the network from high energy physics and elements into sub routines De Laat: Let me try to give then started to work in the and then makes them call- you some context. The first telecom sector in the re- able from application programs

COOK Report: One of these virtualized devices then is a software interface that you use as a s e t o f i n s t r u c t i o n s plugged into another piece of software? You would have many different software descriptions for different devices such that when you activated the description, it can then go to the device, communicate with the operating system of the device and tell it what to do?

De Laat: Exactly. This is his general approach. He

mind around – is that your network becomes just part of your programming environment.

Normally you have your data and your computing and this stupid network that is an unmodifiable “given” that you have to play along with. But here your network is just part of your toolset.

Take the concept of MPI (Message Passing Inter- face). When you have to calculate a huge “do-loop,” then, depending on the calculation, you can divide the work amongst a number of different processors. For works on it with several the network. Your subrou- example one can let ten (PhD) students. He creates tines undergo a series of processors each do one-tenth network objects that are ba- transformations in which of the work; they can work in sically subroutines such that Mathematica tries to make parallel and be ready in one- you can ask them: where are what actually happens more tenth of the original time. you? how busy are you? and closely aligned with your de- MPI is a programming library what can you do? You can sired outcome. that takes care of distributing put a problem in Mathematica the data and instructions that will use these subrou- In some sense what I am de- among the different proces- tines to seek an optimal solu- scribing is a kind of peculiar sors. We see a need and pos- tion for the application. programming language or sibility for a similar frame- programming environment work for networks. This then Within a network, for exam- where your networks are just becomes part of the toolset ple, you can have subrou- subroutines, your data are for creating the necessary tines that describe the topol- more subroutines, your solu- network subroutines that you ogy of the network. Then in tions to solve the problem can use to make a kind of Mathematica you also have are subroutines. You then say automated construction of objects that describe the optimize and solve and it will your desired network usage problem space and the for- then work out the most opti- into a part of your program- mulas needed to solve the mal way to achieve what you ming environment. problem. Next you put brack- want to have done. ets around them and instruct As such the envisioned Mathematica to solve and op- What all this means – and toolset is a programming en- timize the bracketed formula this is the most funda- vironment to optimize the containing those callouts to mental thing to get one’s data and communications

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 16 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 used for your networked ap- ful for a particular purpose If you can manipulate all plications across multiple and twiddle them around and these layers and have also blades of CPUs across the solve your problem doing so vertical and horizontal (wide area) network. Just like ten times as fast in a deter- knowledge in every layer, MPI gives you loop unrolling ministic way. At that mo- you can do the magic and and parallelization you can ment your network be- you get a perfectly inte- also do communications allo- comes part of your pro- grated multilayer hybrid cation and parallelization so gramming language and network that is optimized that your application gets a your problem-solving en- for your application. deterministic network to op- vironment. erate on. That is what we are embark- How the World ing on and these are the first You grab a piece of the com- Changes When You baby steps. My expectation is munications hardware and that, given funding for our optimize it for the problem Give the User his Own work, we are moving forward that you have at hand. Network and we have further research ideas to extend this technol- COOK Report: What do you Now we are taking some ogy. For example this will ex- mean by unrolling in the baby steps down this road. tend our topology, policy, network case? And we have developed some and authorization and forwarding engines and some authentication development. De Laat: Manipulate a net- device engines to manipulate work for your purposes. Sup- tables into packet inspection While your own part of the pose that you have a huge and add headers to packets network functions, the other number of network links. so they can be manipulated, parts may not. Therefore, They are sitting there but mostly in software, in the while you are optimizing your they are not doing anything control and routing portion of network, you will need to for your application. But what this network. take into account the differ- would happen if you could ent usage policies of the net- manipulate them? If you In our group back home we works, that is to say who can could realign them? If you are creating objects for virtu- use the network under what could take them and put alizing and programming conditions. One also needs to them at a different co- wavelength switches and address the cost. Our work location point? Say for exam- photonic devices which di- has several components. A ple, my data center is here, rectly talk to fiber; Micro- policy component; a path but I want my photonics to Electro-Mechanical Systems finding program, an optimiza- go over there. If you have (MEMS) devices that can tion component and a cost the ability to reconfigure connect fibers so that we finding function. these kinds of resources, you have flexibility at the fiber then get the ability to opti- layer. And at the Ethernet Cook Report: if we ever get mize the network tools at layer, we can do similar to this future world where we your disposal. things manipulating Virtual either get rid of the phone Local Area Networks (VLANs). companies or transform them Your application does not We address the Ethernet and we have fiber every- need to use everything that layer and the packet routing where, then does it become is there. You can take just when we need that. possible first to have these those resources that are use- capabilities available to every

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 17 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 fiber connected household in access, you begin to build in pacity. And this is why we are a city like Amsterdam for ex- telecom technology again, especially focusing on ample? But then can we take but on a different level. Here switched photonics and your thinking a step further? it is to enable users, not to avoiding where possible cut up bandwidth and drive much more expensive rout- You will ultimately have a price by making it scarce. ing. In routing the cost fac- vast range of possible things tors are different but also that can be done and you will COOK Report: You have a these services are different. likely be thinking of policies universe of resources that that will authorize appropri- you want to make available Given these considerations, ate users to appropriate lev- under certain conditions to a you need to do things on the els of tasks. You don’t let just universe of users. What I am lowest level of the protocol anyone into a nuclear re- hearing is that under the stack which avails you of the search facility obviously, but right conditions a policy ad- service that you need. If I into things like weather simu- ministrator could work to- have five people using a lation modules might you ward the goal of making printer I put them on the authorize almost anyone who available appropriate re- Ethernet level because oth- was interested and could use sources to appropriate users? erwise it would be a night- the technology in a self mare for the users to get to trained or self-motivated De Laat: Not quite. What the printer. Or should I go to way? Could such users in- you need to give the owner the fiber and say: you need clude all ranges of the public of the resource is the ability to flip that mirror before you from grade schools to high to set his own conditions. If print the document otherwise school graduates to univer- he wants to give them away the printer cannot see your sity graduates to retired peo- for free for example he can document. You go to the ple to people from technical do so. But if you don’t solve layer that is optimal for the schools – you name it? that problem the resource problem at hand. No lower owning people will be very but also no higher. De Laat: I can answer your wary because they will com- question in two ways. The plain that you will also have a COOK Report: C o n s e- resources that I have ex- bunch of people who will quently, at some point in the plained to you certainly rep- want to do illegal movie future if you have someone resent value and we can play copying and that sort of who has developed an appli- around with them as we have thing. cation that could make use of been doing in a fairly small these kinds of network re- research setting. But if we do Now there is another impor- sources, you would be able to not solve the policy problems tant aspect. We want to be put into the hands of that de- of authentication and able to give complete veloper some tools by which authorization, then we will lambdas, in other words he could, if he so chose, con- never successfully roll it out complete colors, to appli- tribute what he had devel- because the people who own cations, therefore, the oped to the kind of “univer- the resources with which cost models that support sal” network administration these tools will enable ex- it will depend on the un- system that would enter it - perimentation will demand derlying technologies like a book - into the library some degree of control over used. The costs involved of this global network where who has access to them. To must be very scalable all the it could become accessible to some extent to determine way up toward very high ca- be designated classes of us-

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 18 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 ers that the developer de- fits into the E-science tech found: data centers, super- sired? transfer conversation we had computers, clusters, visuali- with Kees Neggers two days zation, measurement instru- De Laat: Yes, or for example ago? ments and so on. if you want to do 4 k streaming to Lecture Hall B because I’m looking to continue to On top of that in the light you are about to give a pres- grasp this in more detail from blue bar is a middleware entation there on fluid dy- an understanding of its po- layer that virtualizes all of namics, you tell the system tential impacts on enterprise these resources and with to make the server available business and education and grid technology allows with a lightpath to the display of course I am also wonder- people to build services screen in the lecture hall ing what the business model and applications from the where you will teach. will be because obviously virtualized resources. This nothing comes for free. is pretty much what is al- COOK Report: The number ready happening in indus- of possibilities are unlimited. De Laat: In the diagram try as well; here it is de- above, the gray lattice mesh signed to enhance science. De Laat: For sure. But what is the substrate that connects And on top of everything we we need to do and what I everything together. On the have a number of disciplines was alluding to is to make top of the substrate in the like biomarkers, bio-banking, this process multilayer so brown bar you see a High particle physics, eHumanities, that you can use the layer performance Computing and medical and cognitive sci- appropriate to your task. And Storage and Resource Man- ence, climate research and if you need to work with five agement. This is where all many other disciplines. printers that is one layer; if your computing resources are you want to stream video to a lecture hall that is another layer and if you wanted to use a browser, then I need to be able to direct your resource request to the normal TCP/IP routed layer of the Internet.

What you are looking at is the user interface but what is really important is what is below the interface.

COOK Report: Given then this world in which network tools are virtualized on many different levels. Can we go then to the rest of your hybrid networking presentation Figure 1: The E-Science Tech Transfer World of the Dutch and the extent into which it National Innovation Network

There is also software in here process it here; need to store De Laat: Currently, the gov- that is tailored toward each it there; need to transform it ernment supports Virtual domain of science but still over there and analyze it Laboratory on a project basis. generic to that domain such over elsewhere. At the end of For the last 10 or 12 years that different groups in the this process you will get the government has been discipline can share it. And graphic output, interpret it, funding these kinds of pro- then on top of that you have draw your conclusions and jects. Actually the money to the different domain sci- write a paper. Workflow mid- do this comes from a differ- ences. dleware helps to automate ent Netherlands infrastruc- this basic process of science. ture funding, named ICES/ COOK Report: And when did KIS and FES. The Nether- this start? Treating the GigaPort lands has its own natural gas as Tech Transfer wells. A part of the revenues De Laat: This method of a of this natural gas is put in systems approach to do sci- Infrastructure the mentioned funding ence started about 5 years schemes to support infra- COOK Report: Would you ago with the beginning of the structure. tell me a bit more about what GigaPort and Virtual lab pro- happened to cause the gov- jects. The chart in Figure 1 is Infrastructure, as understood ernment to decide to agree to really a model containing from the point of view of the support this as infrastruc- several projects that are in- history of the natural gas re- ture? terconnected and shown in serve funds, has been high- relationship to one another. ways, railroads, waterways, De Laat: In order to have They were all funded around harbors, and dikes – thus successful science, you need 2004. large scale infrastructure. As a set of tools; supercomput- a part of the basis for the last ers, a data processing grid, a The GigaPort project de- round of funding for GigaPort network, but also common signed the network that is in an article to a national middleware. The common of the gray lattice mesh in the newspaper Jaap van Till, Fe- middleware ensures that not picture. The brown bar is lipe Rodriquez en Erik Huizer every scientist will have to primarily the hardware infra- (*) suggested in august 1997 dig in and reinvent the structure and its direct con- that this money could also be wheels so to speak. This vir- trol software (and virtualiza- used for ICT infrastructure. tual laboratory e-science tion) whereas the blue gray They suggested that for the middleware is regarded as a bar represents primarily cost of one highway crossing kind of tool set that you need software. To do modern sys- the Netherlands could have to provide to the scientists so tem level science you need a the most advanced research that they can do their work. combination of resources, network infrastructure. Then In that sense you should re- instruments, data sources it was decided to start fund- gard the generic e-Science and to enable a group of dif- ing digital highways. (*) services as an infrastructure ferent scientists to work to- http://www.nrc.nl/W2/Nieuws/1997/ for the scientists. The gov- gether to produce a desired 08/21/Med/06.html (in Dutch) ernment realizes this. result. The blue-gray middleware allows you to do your COOK Report: Makes sense. COOK Report: And has science in a workflow envi- I wish we would do some- agreed to support it on an ronment. In other words they thing similar in the USA. But ongoing basis? need to gather data, need to would you go on to talk me through hybrid networking.

and/or send the data to tier- 2 locations and local universities that will process and interpret the data and use the results to do science. Results are then send back to the tier-1’s for archival and access by scientists purposes. [Editor: The slide shown is from Harvey New- man’s September 2008 “deck.”]

Consider the next slide shown immediately below: -- Data Analysis at the Univer- sity of Tokyo. Talk to Profes- sor Kei Hiraki, who has done a number of projects that couple scientific instruments Lab in the United States, to huge databases across Ja- De Laat: Let me do it then German, English, Italian, and pan and use computational from a slightly different point Chinese centers and else- power to do data mining on of view with a November 29 where. The tier-1 centers ei- these databases using very 2007 talk at the University of ther process data themselves high speed networks. Essex called Lambda Grid developments History Present F u t u r e - f o u n d a t http://www.science.uva.nl/~ delaat/talks.html

Lambda Grid Developments and the Need for Hybrid Networking

The first need is illustrated by the LHC Network. You have a scientific instrument that generates data in the amount of Gigabytes per second creating thousands of terabytes of data per year that must be distributed to tier-1 centers for processing. These include NIKHEF in Amsterdam, Fermi

WiVR: a Window into On the demo site using a tion of the panel changes and Virtual Reality tracking system the tablet transmits that rendering in and the viewers position and real time back to the ob- orientation (line of sight) with server holding the panel. The nicknamed “Dead Cat respect to the object were The motivation for this pro- demo” (done at SC04 and at continuously measured. The ject is to test graphical sys- iGrid2005) allows the viewer position information was sent tems and network perform- to explore the internal struc- to the Amsterdam supercom- ance in the context of a ture of a physical dead cat puter where the new slices medical application attempt- (present behind the tablet) are calculated and returned ing to render images in real using virtual technology. The to San Diego to the tablet time. From object was scanned in a CT- display (only a 0.5-second http://www.calit2.net/events scanner in collaboration with delay!). As the person rotates the Academic Medical Center /igrid2005/?p=45 the panel he looks through a at the University of Amster- see also virtual window into the object dam. The data was put in the http://www.science.uva.nl/~ the representation of which is supercomputer to enable mscarpa/wivr/ stored in the computer. The rendering of various slices. panel attached to a high- This has significant implica- This system was put online speed network and correlated tions for surgery and for such that when a request with the eyes of the person other forms of telemedicine. with coordinates of a plane holding it calculates the cor- A scholarly write up has been was received, the system rect rendering of the object published in Future Genera- would render a picture of the being observed as the position Computer Systems vol- requested slice. ume 22 2006 pages 896-900 under the title Highly Interactive Distributed Visualiza- tion.

COOK Report: This can then become a very useful medical diagnostic device?

De Laat: absolutely. You hand out cheap devices for the diagnostic scans while you locate expensive computer, data and rendering infrastructure located elsewhere connected by predictable networks to do the calculations and image processing.

The slide above shows the visualization of cinematic SHD content (3840 by 2160 pixels) on a tiled display setup at the supercomputing center in Amsterdam.

The CineGrid Project

CineGrid”s Mission: To build an interdisciplinary community that is focused on the research, development, and demonstration of networked collaborative tools to enable the production, use and exchange of very- high-quality digital media over photonic networks. For more information see http://www.cinegrid.org/

LOFAR: the Very Long

LOFAR: the Very Long Baseline Interferometer radiotelescope array.

the most hungry namely the C class of users. These are the users of e-science applications where large data flows exist between a very limited numbers of nodes (e.g. from telescopes to correlator, or from CERN to tier- 1 centers). Those flows con- sume the entire capacity of a lightwave on a fiber and, when let loose over the normal routed Internet would disrupt the operation.

The bandwidth for the class A users tend to be DSL speeds, class B ranging upward to Baseline Interferometer radiotele- not use a large amount of Gigabit Ethernet and the scope array is shown in the slide bandwidth while in the B bandwidth needed by the C at bottom of page 23. class the enterprise users group is Gigabit Ethernet and need grid applications, multi- above - including multiples of In the SCARIe project we cast, IP streaming, virtual that 10 Gb and 100 Gb. What use lightpaths and photonic organizations, mostly com- we need to realize from this networks to collect all the posed of LANs. They need is that one size does not fit data from the radio tele- VPN services and full Internet all. We need to work with a scopes to put it into a distrib- uplink routing. Then at the combination of solutions. uted software correlator. The far end of the bandwidth are result is as if one is observing the sky with a very large telescope with extraordinary angular resolution.

What These Large Scale Projects Teach Us About User Populations

And below is the ABC user classification slide. It shows how strongly asymmetric the amount of data used by a diverse user population is where the vast majority of people -- your lightweight A class users – need full Inter- net routing but to do mainly browsing and e-mail and do

For the whole slate of applications on the Internet one will need a combination of the technologies described above. Given the amount of bandwidth that it enables, photonic technology is getting very cheap.

If we begin to switch light waves, we avoid latency and congestion and find that gig Ethernet and above is usually an adequate carrier protocol and that applications can have control over network behaviour that would be very difficult in a routed network. COOK Report: in other in a fiber toward the outgoing words the evolving Internet fibers. Using Photonics also COOK Report: so what you needs to be a hybridized sys- takes less energy. But of are saying is that if you are tem that can satisfy all three course if you route you can going to invest money in classes of users enabling go specifically where you networks you must look at transitions from one class to wish to within the global your investment in a very another where necessary? Internet. The goal is to de- holistic all-inclusive way. And sign your network in a way that the appropriate use of De Laat: Yes, this was one such that each packet is your resources would be to of the mayor findings in the given the minimal service match the delivering tech- GigaPort project and this is that it needs to do reach its nologies to the needs of the presented in the slide above destination and no more. three classes of users leaving “Towards Hybrid Network- ing.”

The cost of photonic equipment has come way down to where it is about 10% of the cost of switching which in turn is about 10% of the cost of full routing. Obviously the services are also different. Routers “know” the topology of the entire Internet and can route per packet, switches using layer 2 information are only aware of the LAN, and the fiber switches only forward entire colors or all light

hybrid networks as the logical as possible and go as far as is happening and can see outcome of this matching of possible at the lowest layer the benefits of not always economic investment with before you need to bounce up working at the highest technology? higher in the stack in order to and most expensive layer do what is necessary to get -- namely the routed De Laat: Yes just as in com- your packets or photons as layer. One needs to make putation where one needs to the case may be to their final sure that your users un- balance grid and supercom- destination. derstand the benefits of puting technology. That is an operating in different lay- understanding that we still COOK Report: Then what ers of their optical net- need to spread. In short we started out talking about work. The users need to there is not one solution that is how to build a network- grasp that if they endow fits everything. You need a ing environment that their applications with the combination of solutions for combines this use of ability to intelligently your network transport as equipment, technology, traverse the layers they well. and energy in such a way can open up all manner of to give the user applica- increased possibilities to How Low Can You tions that meet his de- better performance at less Go? mands at the minimal cost. To do this you effec- necessary cost? tively need a kind of “wall To answer this problem you connector” to your optical must ask how low in your De Laat: Yes. But to do network represented by the stack of network layers can this one must expose the cream -yellow box on the left you go? You want to go as network to the users so of the slide above in such a low in the service layer stack that they understand what way that the applications can enter the network at the layer on which they can operate most effectively. What I am talking about here is effectively a control or service plane for the network.

The Global Fiber Infrastructure

The chart at the top of the next page represents the GLIF or Global Lambda Inte- grated Facility. This is the global infrastructure of fiber that we are using for developing the hybrid network tools I am explaining to you.

Also in the next two slides I Your Network Application Must Use the Appropriate Transport Layer present the situation around

the Netherlands, NetherLight and Amsterdam. First and to the left you can see the connectivity and, in the second graph at the top of the next page and overlaid on the connectivity map, the activities that flourish around NetherLight in Amsterdam.

On page 28 you will see the map of dark fiber in the Netherlands. SURFnet has some 8000 km of fiber - an infrastructure that is larger in length than the railway system of the nation.

The “map” above shows the four major research areas made possible by the Dutch infrastructure: Visualization, Data Storage and manipulation, grid and cloud computing and supercomputing. The map below shows the SURFnet ﬁber infrastructure.

COOK Report: of the 160 roughly how many are enter- De Laat: perhaps about four connections of SURFnet prise research centers? or five

resources and to create multiple overlay networks, each with a different logical topology. The novelty of this project is that it does give this flexibility directly to the applications by allowing them to choose the logical interconnection topology in real time, ultimately with subsecond switching times. The novelty of StarPlane is that the change of topology is implemented with photonics; namely Wavelength Selective Switches. So StarPlane rear- ranges colors in the wide area dark fiber infrastructure. http://www.starplane.org/

supercomputer DAS-3. Hy- The Signiﬁcance of We do this experiment with brid optical networks such as SURFnet6 as a five distinct clusters at 4 dif- SURFnet6 allow network us- ferent Universities - hence Photonic Network ers to partition the network

This is the subnetwork map of SURFnet’s fiber rings where they can do 10, 40 or 100 Gb per lambda and transport up to 72 lambdas per fiber.

Together with SURFnet we have implemented the Star- Plane project with major contributions from Nortel. Its purpose is to allow part of the photonic network infrastructure of SURFnet6 to be manipulated by Grid applications to optimize the performance of specific e- Science applications.

StarPlane uses the physical infrastructure provided by SURFnet6 and the distributed

where every color goes to a different angle. They strike micro-mirrors that reflect back on the correct angle the color that you want to have put on the output fiber. In this sense you can say I want red from the bottom fiber and blue from the top fiber to go in the middle fiber. This means then that blue from the bottom fiber cannot go in the same place because the mirror is already occupied. As a result you gain the ability to mix and match colors.

Such a photonic device costs about 10,000 dollars, obvi- the logical view of a Pentagon A Photonic Switch ously more if it is packaged shaped diagram. The de- and enabled with a controller ployed application signals to The slide below [Module Op- with software. It allows flip- the NOC what kind of connec- eration] shows how our ping around some 80 differ- tivity it wants. wavelength selector switch ent wavelengths in five or works. The light comes in and eight fibers. With plenty of lambdas and strikes a diffraction grating the ability to switch new ones on demand, we can have here what I like to call QOS in a nonde- structive way. If you have some new users call up some new colors, then you map those new users to their new colors and everyone should be happy in contrast to differentiated priorities schemes, where some users get less priority than others.

COOK Report: Is this light- pologies. You optimize the not the same as lightpath path switching? topology of the network for switching? where you see the most load. De Laat: Yes and no. The If you see that you need a lot De Laat: If you call a light- term lightpath is used in a of capacity to a certain site path a true lambda, then it is much broader sense. This is a you create a tree structure of the same. But people also very special case. This is lambdas to go to that site. call VLANs, from ingress to thrue photonic switching. You And if in the next round you egress, “lightpaths.” In this are really switching colors. In see that more next neighbor sense people pollute the other words lambdas which communication is needed, name “lightpath.” It is not you mix and match. If you you make a ring topology. light and it’s hardly a path. build an architecture com- You can do this by flipping posed of a number of these the mirrors on-the-fly to Below is a slide describing switches, you can send colors change the underlying topol- how the dispersion compen- anywhere you want them to ogy of your photonic net- sating modem from Nortel go without having the traffic work. Since you do this in operates. But this may be too to go to the electric domain. photonics you do not need in technical for this discussion. OEO device sitting around to Basically one figures out what Now if you want to do OEO make the changes. the distortion of the signal that is optical to electrical will be after transport conversion and back to opti- COOK Report: But in the through the fiber. Then you cal you need colored lasers case of the module operation pre-distort your signal in the that cost you many 10ths of slide above, we are talking opposite way such that when thousands per laser. That is about a lambda as an optical that pre-distorded signal gets very expensive. A WSS de- Lightpath that can be redi- distorted through transporta- vice that can do hundreds of rected on the fly? How is this tion it is normal again. You colors photonic switching costs you also 10ths of thousands but you only need one. This is what makes photonic operation cheap.

COOK Report: Take me up a couple of levels. Here you are down at the very bottom ground floor. The purpose of doing this is that once you have done that you can mix and match colors and get your photons where you want to send them much less expensively?

De Laat: Yes, think about what we have shown in this demonstration where you play around with these to-

Austin (TX). The example is from SC06.

We created an ontology to describe the components that you find in these kinds of networks.

Based on this ontology, we have schemas to describe the placement of routers, switches, interfaces, links and fibers and the framing that is used to transport data. Given that we have this ontology, we can use Web 2.0 technology to reason about it - just like friends of a friend’s network where you pre-distort your signal in domains is the description of say Johnny is a friend of such a way that the fiber’s the topologies for pathfind- Carla and Carla knows distortions will cancel out the ing. The RDF network de- George and therefore there is pre-distortion. scription language is there to a path from Johnny to master and describe the George. You can take these Within grids you usually co- kinds of infrastructure sets relationships and turn them schedule compute and data that we, for example, see in a kind of map. resources but now you can from SciNet here at SC08 in also co-schedule your lambdas within the grid. This gives you a new playing field that is shown in the GRID Co- Scheduling Problem Space slide above. You get very constant behavior over the lamdas. That is actually trivial because there is nothing in the path which can make it non constant. Your round-trip times are always exactly the same.

Resource Description Framework Language

One of the major problems when dealing with multi layer networks spanning many

to this interface is connected to that interface by a fiber.

This slide at the bottom of this page describes the modeling process for an Ethernet and a fiber layer.

These descriptions permit us to point to other domains. Consequently you can have a description of your network and point to a description of neighbor networks, thus creating a web of descriptions. Each network maintains the description of their own network and point to the others where they connect.

scribe a friend of a friend in We do not need one “master- triplet descriptions. We figured out that you could of-all networks” description. do the same in networks and You can just interlink it as a use Web technology to de- The statement Carla is a Web object. This allows you friend of George is equivalent to do path finding for these descriptions.

COOK Report: And conceptually this is down in the engine room of the path finding process for which the light table on the show floor represents the graphical user interface?

De Laat: Yes. We have tools for the network description language. We showcased that by describing SURFnet in resource description framework slide at the top of the next page.

COOK Report: An NDL file is what?

De Laat: Network Description Lan- guage of which this is an example.

And in the slide above presents tools to generate a NDL file for a setup and validate the syntactical correctness of an NDL file. The slide below shows an example of the multilayer features that can be described in NDL.

The slide below is

a visualization of SURFnet6 based on a NDL description. Al- though the visualization has lots of room for improvement, it gives confidence that the model works.

Network De- scription Language as Routing Pro-

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 35 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 tocol quires a loop to allocate a Net and Ca-Net to StarLight gigabit lightpath from source and then via MANLAN to Am- In the next two slides we to destination. sterdam? demonstrate the multi domain multi layer routing COOK Report: So in the first De Laat: That is correct. The power of the model in a slide below you can not go capacity of CA-Net to Star- Light in this mock-up example is insufficient. So the traffic has to travel via MANLAN. It must go to StarLight to be translated in the correct data-framing that the destination can understand. But when it gets to StarLigth it cannot travel back to MAN- LAN since the remaining capacity is now also too low, but since the new framing occupies a bit less capacity the traffic can now flow to CA-Net and then to Amster- dam via MANLAN. Hence pathfinding results in the loop as shown in the slide below.

COOK Report: This is like network state information?

De Laat: Yes and it can be described in NDL that is Net- work Description Language. I think that there is no other operational routing protocol in the world that would permit you to be able to figure out a loop like this.. But we can do this.

And on the top of the next page is a demonstration where the protocol goes through the steps of the state machine.

mock-up network that re- directly from Québec to Ca- By tracing the course of the Green line from the left and

content, projectors, tiled displays and transcoding services, then we are able to ask a high-level question such as: “show me video with this content and show it to me on my tiled display.”

If you then find out that the video is resident on this server, that it according to its metadata needs to be trans- coded to be viewable on a tiled display and what network capacities are needed, then we can request the necessary lightpaths and resources to make it happen. Given the relationships in RDF we are assured that the the right you can see how the resources are compatible. It network protocol tries all We want to extend NDL to is just a matter of pathfinding possibilities and finds its way describe other types of infra- in semantic space that trans- through the maze. We are structure. If you are able to lates to co-allocation of com- submitting NDL to standardi- create ontologies and sche- patible resources in real zation organizations. mas that describe storage, space.

COOK Report: Who else in the world is doing anything like this?

De Laat: Not many groups and those that work on NML are basically working with us. The RDF wikipedia entry says “The Resource Description Framework (RDF) is a family of World Wide Web Consor- tium (W3C) specifications, originally designed as a metadata data model, which has come to be used as a general method of modeling information through a variety of syn- tax formats.”

Terabit Networking

You may ask yourself what constitutes a terabit per second network? A laboratory such as CALIT2 has 8000 Gb drops. Does that constitute an 8 Tbit/s LAN? At the Uni- versity of Amsterdam we have 2000 1 Gbit/s drops. Does that make a two Tbit/s lan? I don’t think so because it depends on what you add ond! Of course some of the COOK Report: Your point is up. If you look at an 64 core capacity on those chips is it’s all a matter of the context Intel processor and you cut it there just to let the cores talk in which you do your think- in two, the left and right half with each other. ing? communicate at 8 Tbs per sec-

De Laat: Correct. I consider it to be a Tbit/s network if all To say it in a different way; You look at it as though it that capacity can be used by the displays for an OptiPortal were a bunch of small Lego one application for its pur- tiled panel are useless with- building blocks. You say if I poses. For that one needs out the middleware that put this object in here I need programmable networks. So makes the tiles work as one bandwidth to talk to the core we asked the question of big display. and then I must be able to what constitutes a terabit write it to a disk array. You network in a different way. What we are doing with our need to be able to coordinate Think back to teraflop com- resource description frame- and allocate discrete compo- puting. People started out work is to drive the network nents so they work together here by filling up rooms full to do something collective on in an organized and coopera- of PCs and interconnecting behalf of your application. tive way. them. But doing this doesn’t The collectivity makes it a Tb make it a teraflop computer. resource for the benefit of The next slide explains the your application. But to programmability of the net- What makes it teraflop com- achieve this collectivity you works by embedding the puting is MPI or Globus. It is need programmability and network elements in func- the middleware. Such mid- middleware. dleware is necessary to ensure that an application like fluid d y n a m i c s modeling can har- vest 8 t e r a f l o p s worth of computing power for its own purposes. You need to have middleware to make this technology work for you at the teraflop or t e r a b i t thinking level.

Finally, in dealing with ICT infrastructure, power con- sumption is a big issue. The more switching you can do in the photonic domain, the more power you can save.

For every dollar of equipment you spend you can count on spending about one dollar for electricity to power the equipment during its life time. That is true for computing. Routers are so expensive that the power is relatively a minor cost factor. Still, if you put 80 colors into a router you need an expensive router interface for each color and tions that can be used in data at the compute ele- Mathematica formulas. ments in time for processing. The slide above shows that The use of Mathematica al- applications of the future lows to optimize the network need to be aware of all the on the fly for the problem it layers in the network. has to solve; i.e. getting the

Cees de Laat at SC08 in Austin

each of those interfaces consumes 200 Watts.

Editor’s Note: The prototype discussed below was developed by Rudolf Strijkers under supervision of Prof. Dr. Robert Meijer as part of his phd research on next-generation Inter- net architecture. Mem- bers of the team that developed the demo were Mihai Cristea (post-doc), Laurence Muller (scientific pro- grammer) and Robert Belleman (head of UvA visualization lab). My Interview with Rudolf was conducted on No- vember 19.

COOK Report: Cees de Laat and I’ve been Figure 1. Interactive Networks setup at Super Computing 2008, down in the engine room Austin, TX. talking about what Strijkers: This is the first makes these user controlled According to what you were prototype of what we are and application controlled telling me the goal of what calling “Interactive Net- switched light path networks we are talking about is to works”. In interactive net- possible. Now we are going have this kind of software on works humans become an to look at a prototype of a an ordinary researcher’s integral part of the control user interface with a multi- screen three or four or five system to manage the next- touch-sensitive screen that years from now. At that generation of programmable allows a user to tap on the point the user should be able networks and Grids. The tools he wishes to select and to use it to control most any main design principle is this; with his finger to draw the application in collaborative by virtualizing the configur- paths he wishes to activate environment. Would you take able and programmable nodes on a programmable me on a guided tour of what properties of network ele- network. Multiple researchers it is and what it does and ments as software objects, can use the multi touch inter- how it works? any aspect of a network in- face at the same time. frastructure can be manipu-

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 41 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 lated from computer pro- nection. This way, the Mac over 700 nodes within sec- grams. What we show here is mini’s are part of the subnet. onds. The neighbors you dis- an implementation of an in- cover at Ethernet level will teractive control system con- When the network boots and look like fully connected at cept for user programmable the nodes come up they will the IP level. For example, if networks, which applies the connect to controller. The you interconnect the three architectural concepts we controller is programmed in computers with a switch, it have developed in our re- such a way that it will send a will always look as though search. out a discovery request when each computer can reach a new node connects. Each each other directly. That’s The network you see here is node will try to discover its why it looks like three fully the current set up of our test neighbors using ARP to scan connected networks here and bed located at the University the whole subnet for hosts. one large fully interconnected of Amsterdam. These icons Since it will discover every- network over there; the video represent the network’s ele- thing in the subnet, we also screens are just one hop ments and network structure find all the nodes in the data- away. visualized at IP level. I can center. We currently only tell you a little bit about the display the programmable Once we have discovered the actual infrastructure. There nodes, but the whole discov- networks you can also see are currently twenty nodes, ered network can be dis- the result over there in the interconnected by four sub- played too. Mathematica interface on the nets to create an operation- fifth screen. The Mathematica ally interesting topology. At SC08 we forgot to turn off interface has access to the Three of these subnets are in WiFi to Scinet on the Mac same information and sup- separate virtual machine en- mini’s once and discovered ports the same network ma- vironments of VMware, called ESX servers, that also contain four virtual machines each. ESX is essentially a container for virtual machines. The VMware management environment enables us to create, clone, and remove virtual machines. It also enables creation and manipulation of complete virtual network environments.

The virtual subnets and machines are connected to a physical subnet, which also contains two physical nodes. Then we have the four Mac Mini’s here in the booth, which are directly connected Figure 2. Network visualization after discovery. The dashed lines, when to the physical switch in Am- monitoring is enabled will indicate bandwidth (line width), delay (percentage sterdam with a gigabit con- of dashing) and jitter (randomize in dashing).

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 42 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 nipulations as the touch ta- way a node never receives a video stream appear on the ble. We will come back to stream, except when we run screen. this aspect later. our special expressions to capture the traffic. COOK Report: The blue line COOK Report: What do Underneath the node you can is traced with the finger? these icons represent? see a small graph, which will [See Figure 4 on page 44 be- displays CPU load measure- low.] Strijkers: The icons repre- ments. A button on the top sent the type of modes or right of the touch table will Strijkers: Yes. If we make a functionality that each node enable or disable the CPU path, what will happen, we offers. We currently have load measurements in the send a request to the control- three modes. 1. A producer: network. The real-time load ler asking to create the path this node contains streaming information for a selected that we just dragged with our video content and is visual- node will be displayed in this finger. The controller will ized as a green-circled arrow. graph. When the load be- send the request to a com- It can also route traffic. 2. A comes larger than 1, the piler to generate the com- consumer: such a node is nodes will light up red, alert- mands and forwarding ex- connected to a streaming ing the operator that the pressions for provisioning the video client and can display node is under stress. nodes. The results will be the streamed content. A play passed on to a transaction button in a screen shows it. Now we can decide to make a monitoring, which executes a 3. A router: The sole purpose path. Then we go into the distributed transaction to of these nodes is to route path creation mode, and then load the commands and ex- traffic and blue-circled arrows you can decide to trace a pressions on the nodes. If demonstrate it. path, from a producer node loading of one of the expres- to a router to a screen, for sions or commands fails, the At this point we can look at example. And you see the whole transaction will be what kind of videos may be streaming in the network. You can push with your finger on a producer node and a window opens on the touch table that gives you a preview of the video. By the way, the movies we currently have are: Big Buck Bunny, Elephants Dream, and two Cinegrid demo movies. The first two are actually made in the Netherlands as part of the Orange Open Movie Pro- ject and the son of Cees de Laat made one of the Cine- Grid movies. All the movies stream continuously stream in high definition, but to un- Figure 3. By touching a producer node you will see a preview of the known destination IPs. This video stream. The small graph underneath shows real-time CPU load measurements.

Strijkers: Yes, you will. The interface is very simple considering the capabilities. Ac- tually, what we show here is a showcase for how we envi- sion the management of the next-generation networks. So, the capabilities of our experimental programmable network are one step further than what you can do with modern networks now. For example, I can show you how we can draw a path with a loop. Have you ever seen a Figure 4. Simply drag a line with a finger to create a path. path with a loop in a network? rolled back. So, whatever node and drag it to a screen happens in inserting or re- over another route. And then COOK Report: So, in other moving requests, the network you can see the movies words, that path that you will always be in a consistent streaming on two computers. just set will send data back state. When drawing a path, and forth two or three times? the touch table will also COOK Report: With this ta- automatically select an un- ble, and if somebody will Strijkers: Yes, the packets used color and this color will show me, I’m sure I will learn will ping-pong back and forth be used for lifetime of a path the basics of it pretty quick, before being routed to the to identify the stream. When right? screen. Normally, loops in a tracing the traffic on IP level you will actually see the color codes in the IP packets.

Once we have a path we can also select it and extend it to a multicast path. This can be done by dragging a new route starting from any node of the path chosen for the extension. It’s as easy as that.

COOK Report: You are taking this content and sending it to a second screen?

Strijkers: Exactly. I only Figure 5. Selecting and extending an existing path from a node have to touch the path on a creates multicast paths.

colleague Leon Gommans did a lot of work on this subject and we have extended his ideas to programmable networks. In the demo, the compiler detects the loop and creates expressions that change the colour of a token to a different shade. This means that at every hop the token is rewrit- ten and will flow in another streamline graph. For example the colour still remains blue to the application, but in reality the network uses the shades to maintain state.

So, I have shown how to Figure 6. Creating a loop. Unicast paths can be routed any way a user make a unicast path, how we likes, whether it contains one or more loops or crossings. The compiler will detect and generate the correct expressions. made a multicast path, and how we made a path with a loop. Now I can show you a network are bad, because only after my pizza arrives’, little bit what happens inside routers have no way of de- but also to bind network be- the node. tecting if a certain packet al- haviour of communities or ready came by or not. distributed applications in COOK Report: OK grid networks. Our former To achieve loops in a programmable network, you could use special programs with counters to detect loop- ing packets. But, we enhanced IP a little; we put a token inside the packet, in the IP option field to be more precise. This token is not necessary, but it allows us to white list or identify packets or streams uniquely. The token enables us to bind network behaviour to traffic that is not in any sense connected to the protocols used. An example of such a binding is ‘I want a good quality video Figure 7. Changing path creation mode. The button with the question connection to my TV, but mark shows the discovered topology including non-programmable nodes.

Strijkers: For this we need every node. This flow graph mit is actually the output to switch to a different mode. will describe how the traffic function of the linux kernel. This mode disables dragging flows from the input ports to So, if we send a packet there, of paths and allows us to in- the output ports. it will be routed and sent to teract with the nodes and the correct host. We have edges of the network. When I COOK Report: And the out- made a special modification, double tap on a node, you put port is skb in the red cir- were we have full control will see what happened in the cles? over the traffic flow. This first node when we made the filter is tbs. It’s called the loop. Strijkers: Yes token based switch. What it does is, it looks at the packet COOK Report: The large COOK Report: And the tbs, and says, it’s a blue packet, black circle indicates you what does that stand for? and I’m a blue graph, so I zoomed into the node? accept the packet. If I would Strijkers: Let’s start at the be a red graph I wouldn’t ac- Strijkers: Yes. When we first filter first. Netfilter is a cept the blue packet. In other make a path, like a loop for library in the linux kernel, words, it accepts or drops example, the request is sent which captures all the data packets based on their token. to a controller. This controller from the networking stack at This allows us to create runs the request through the specific points. When not application-specific flow compiler, which checks if the used, traffic would normally graphs for tokenized streams. nodes are available, how they go through the normal net- are connected, if the source working stack of Linux. But, When the packet is accepted, is a video stream and if the what we do is we capture the it goes on to the tb and the destination is connected to a traffic at Netfilter input and tb filter tears off the token. screen. The compiler will force it to go through the Why would we tear of the to- generate a flow graph for flow graph. And, skb_trans- ken? Because if we send the message to the Mac mini, it would have no clue that we did all kind of weird stuff with the packet. It just looks like any other packet. And here is the magic; we have a filter that rewrites the IP destination of the packet to route it to go there. The library that allows us to insert/remove these filter expressions at run-time in the kernel is called Streamline. It was developed at the Vrije Univer- siteit of Amsterdam and our colleague Mihai Cristea and I worked closely with its developer Willem de Bruijn to make it suitable for our pur- Figure 8. Streamline ﬂow graph currently loaded in a node. Note that due pose. to the picture contrast the connecting lines are barely visible.

value) So, you just saw me modifying the flow dropping rate in real-time. Now we only implemented user interface support for a sampler, but you can insert any type of filter yourself or even write your own.

This is a powerful tool to exert very fine-grained control over traffic. For example you could add or manipulate rate limiter filters, which would allow very precise traffic shaping. The operator at the touch table could manually control the traffic shaping, Figure 9. Modifying the sampling rate of a ﬂow. The extreme left shows but it is also possible pro- the manipulated stream, and the screen immediately to its right shows grammatically. I can show an unmodiﬁed stream that is also part of the same multicast tree. this later on. Strijkers: Not exactly, it COOK Report: When you adds a filter to the run-time Strijkers: And we can say, get a green circle with a plus, expression at a certain place. hey, we change the sampler what does that indicate? Keep in mind that the actual to 80 percent. You will see code is running in the kernel. that the screen gets gradu- Strijkers: The compiler ally better. It’s because of automatically generates What actually happens is that the encoding that it will do these expressions, the dis- this request goes to the spe- weird stuff. But, we can also tributed transaction proces- cific node, it picks out the remove it with the minus. It sor executes a two-phase manipulated expression, will go streamline again, it commit on all the nodes and plugs the sampler in and puts will remove the sampler and inputs them in Streamline. it back into Streamline. And you will see the stream turn- After the transaction is com- you can see it, because your ing back to normal. plete, we can zoom into the whole video stream goes node and modify the expres- nuts. On the left screen we COOK Report: Impressive! sion that resulted from the see the video of a multicast compilation process and that branch with a sampler and on Running Mathematica is currently running in the right the unmodified Streamline. Just by touching stream of the other branch. Strijkers: So, that’s the the plus button. touch table part. The other The image is distorted be- part is we can enable COOK Report: That gets you cause we throw away some throughput measurements, to a different interface, or? packets. Right now it throws so then the controller will ask away 50 percent of the pack- to the network to continu- ets. (Modifies the sampler ously return throughput

ble network and now Mathe- matica is part of the programmable network and is able to receive measurements or manipulate paths just like the touch table.

COOK Report: You can see the orange yellow cone changing shape, and it looks almost like the visualisation of a pumping heart. That’s my metaphor.

Strijkers: It’s a 3D contour plot of the real-time through- Figure 10. A closer view on the sampler modiﬁcation interface. The put in the network, so the sampling percentage is changed with the slider. analogy with the pumping heart is quite accurate. The measurements. And you can you to write programs and contour plot shows that we see what happens over there. dynamic visualizations. I im- can now directly apply the plemented a Mathematica large collection of Mathe- COOK Report: You’re over interface to the programma- matica libraries to visualize at Mathematica there

Strijkers: Yeah, Mathematica is a scientific computing environment. It allows you to do interactive calculations. Mathematica contains a large library for statistical analysis, graph theory and so on. And by the way, today they released version 7, which includes out-of-the- box support for parallel computing. Mathematica also is a powerful symbolic Figure 11. Interactive Networks in Mathematica. The left window shows the 3D contour language that allows plot of the real-time throughput in the network. The right window shows the current topology of our programmable test bed.

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 48 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 and program networks inter- actively while using real-time measurements. On the touch work adaptation in a user- manage programmable net- table we do graph layouts by friendly environment. works. But, it is not limited to hand for example; Mathe- this case only. Amongst matica can calculate the lay- For example, now we can other things, we would out automatically, which you say, certain paths should like to incorporate light- can see in the other window. avoid busy parts of the path management and in- To create the dynamic plot network. By using only clude other resources, that is shown on the screen, standard functions in such as storage or virtual a graph layout algorithm of Mathematica, it is already machine management. We Mathematica determines the possible to write a simple hope that on a longer term form of the surface. Then the program that uses the we can do trials on a throughput measurements real-time throughput in- larger scale, for example, make up the values of the z- formation to continuously by integrating our solution axis. reroute one or more paths with Internet2, ARGIA and that avoid busy parts of other research networks With the information the network. and Grids. Eventually, we given by the network, and hope to see our work ap- currently we support con- COOK Report: In three plied in collaborative con- tinuous measurement of years time, this software and trol room environments delay, jitter, bandwidth capability should be on every for monitoring and control and throughput, one or researcher’s workstation? of complex and large- more operators at once scale systems. can write programs that Strijkers: We certainly hope automate decision- so. The demo we showed to- making. This opens the day illustrates a novel way to way for automated net

Conclusion Where Is All This Headed?

The technology and projects sponded: “In the Netherlands is too distant from the we have just described are we now have all organiza- core competence of the not destined to remain as tions connected to SURFnet6 researchers involved, and high-end toys. With increas- on SURFnet leased fiber, there is always a reluc- ing improvements in hard- most of which is based on 15 tance to spend money for ware, software and band- year IRUs. One third of them common infrastructures width there is no reason that are using lightpath services with uncertain individual these capabilities cannot be today. Since last December paybacks. Generally it is headed for enterprise, educa- users can now dynamically possible to get a project tion and consumer markets. provision and reserve light- manager to recognize the Of course predicting any paths via a simple web inter- desirability or even the roadmap with precision is dif- face. We plan to make the necessity of the use of ficult. The outcome will be network even more optical lightpaths but at the same determined by the intelli- and “green” and will study time it is difficult to get it gence that underlies compet- how we can benefit from the accepted on a higher level ing policy views. emerging “Next Generation and implemented in the Ethernet” to bring more project execution, often As the interviews in this issue granularity and scalability to because it does not fit into make clear, over the past the lightpath provisioning.” the original funding three years, these technolo- scheme.” gies have been successfully “At the application level we used in international scientific notice a rapid uptake by the And last but not least, as experiments and collabora- ICT departments for their proposed by ICTRegie in the tions. Also as Kees Neggers own use. Also the big science report “Towards a competi- points out: not just for re- users like particle physics and tive ICT infrastructure for search. In the just concluded astronomy, who created the scientific research in the GigaPort project they have first demands for lightpaths Netherlands” we will have to also successfully been used and have their own dedicated increase the coherence be- by ICT Departments to create ICT know how and support, tween organizations, activi- optical private networks have successfully integrated ties and the components of (OPNs). These OPNs link dis- lightpaths and OPNs to sup- the ICT infrastructure. locations of universities and port the international collabo- research institutes together, rations in their fields. But In other words an emphasis making centrally provided outside these communities in standardization of optical high quality e-learning facili- the uptake is not as good.” network design and project ties and collaboration tools coordination among partici- available at all locations. “We have concluded that pants helps to ensure that more selective promotion and the technology is actually When I asked Kees what next support is needed to attract used. For the researchers steps he foresaw, he re- users from other sectors. ICT easily used applications tools

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 50 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 and training and support for control and management pro- processing of that bill and their use is critical. Ed Seidel tocols. 10% went in the costs of the as Director of the Office of actual call. So what can we Cyber Infrastructure at NSF 5. Virtualization using WEB learn from that? :-) made this point strongly in 2.0 technology to enable the his BOF presentation in Aus- operation of the programma- tin saying that the High per- ble network paradigm across For Further formance community in the domains. Investigation - US must focus on outreach the OptiPuter and application tool develop- 6. Policy and authorization ment. frameworks. Two issues of Future Genera- tion Computer Systems On the research end Cees de 7. Interlinking different NREN (FGCS) are worthwhile read- Laat pointed out that major hybrid networks such as ing. The first is Volume 25 goals include: Internet2, GN2/3, Phospho- issue 2 Feb 2009 entirely de- rus test bed, G-Lambda and voted to the OptiPuter Global 1. Development of common others for dynamic service Collaboratory and FGCS, interfaces for application setup. Volume 22, Issue Number 8, software to interact with the October 2006 on “iGrid new networks so that distrib- There is no longer anything 2005: The Global Lambda uted applications can use the that is hindering static serv- Integrated facility.” special features of the net- ice setup. So many pieces of works. Standardization ef- what we developed in the See also information about forts are needed here and are past seven years can now be Optiputer found at underway in a number of implemented by everyone http://www.optiputer.net/: standards organizations. today. “The OptIPuter, so named for 2. Development of a topology Cees also put it very well its use of Optical networking, distribution infrastructure for when he wrote: making this Internet Protocol, computer pathfinding through these broadly available will require storage, processing and visu- networks. Here also efforts a shift at the provider’s busi- alization technologies, is an are underway. ness models where they open envisioned infrastructure that up as much as possible the will tightly couple computa- 3. Greening the network by infrastructure and offer it to tional resources over parallel keeping traffic as long as the customers. It seems that optical networks using the IP possible in the photonic do- the current model is to cut up communication mechanism. main. This includes light service in small slices and sell passing through other do- that for the highest possible The OptIPuter exploits a new mains. Technically this should price. The problem is that the world in which the central be possible; culturally it is a infrastructure needed for that architectural element is opti- challenge. cutting up is expensive and cal networking, not comput- becomes another item for ers - creating "supernet- 4. Study using simulations of which customers must pay. works". This paradigm shift the scaling properties of such Some-one mentioned once to requires large-scale networks with respect to ag- me that 90% of the bill of applications-driven, system gregation strategies, and telephony was for the meter- experiments and a broad ing instrumentation and multidisciplinary team to un-

derstand and develop innova- end computing, storage and works in the Netherlands and tive solutions for a visualization resources.” The the creation of a virtual labo- "LambdaGrid" world. The goal OptiPuter has its own wiki. ratory for e-science to further of this new architecture is to http://wiki.optiputer.net/opti technology transfer between enable scientists who are portal/index.php/Main_Page Dutch universities and indus- generating terabytes and pe- try, one of the things that tabytes of data to interac- This is fascinating since has struck me most strongly tively visualize, analyze, and among other things it teaches is the very clearly defined correlate their data from mul- everyone how to build their approach taken by the Minis- tiple storage sites connected own OptiPortals. Everything try of Economic Affairs, min- to optical networks. is open source and not istry of education, Science stratospherically expensive. and Culture and the Dutch And then look at Optiplanet: equivalent of the national http://www.evl.uic.edu/caver There are approximately 40 science foundation to estab- n/optiplanet/ “From 2002- OptiPortals installed around lish and organization known 2008, a global team of re- the world. The illustration as ICTRegie. searchers and networking above shows 12 of the early engineers has been building ones. Reading the five year plan of the OptIPuter, a National Sci- the Dutch ICT Research and ence Foundation-funded ini- Itʼs the Process and Innovation Authority shows tiative to dynamically config- the Planning the development of a very ure a distributed computa- sharply focused way of ap- tional facility, where the opti- In the process of putting to- proaching the expenditure of cal network becomes the gether this issue on the de- government funding for re- “backplane” connecting high- velopment of lightpaths net- search and innovation on be-

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 52 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 half of the interests of society achieve a balanced portfolio processes, transport and stor- at-large. The plan is a 41 of investments in ICT re- age, as well as carefully coordi- page pdf downloadable at search and innovation by nated cooperation between all http://www.ictregie.nl/index. balanced cross disciplinary links in the chain from producer to consumer php?pageId=6&pubId=15 work in four market sectors, (Go to this page and click on four social domains and four Water - the greater computa- the link “downloads.”) application areas within ICT. tional capacity now available enables mathematical models Although I quoted this in the Economic Domains - with a growing degree of inte- introduction on page 2 Here the first domain is: gration to be produced. Com- above, it bears repeating. bined with data-gathering and The decision to form ICTRe- Hi-tech systems and ma- interpretation techniques, this gie was based on the under- terials - the field of high-tech opens the way to the develop- standing that “within the systems and materials is con- ment of new control instru- complex field of ICT, with fined to the high-tech manufac- ments, perhaps linking meteor- its many research disci- turing industry, i.e. machinery ology with water management, plines on the supply side, and the material-reliant bio, or in the form of innovative in- countless application do- pharmaceutical and chemical spection methods for waterways mains on the demand side sectors. Modern technical and water barriers. and a large number of or- equipment (such as production ganizations throughout, a machinery, cars, baggage han- Social Domains - dling systems and high-speed purely top-down style of Heathcare - (in three parts) screening systems) is controlled direction is unlikely to be by embedded software. . . . . - effective. The healthcare client: the Smart Health Surroundings Creative Industry - In this theme is concerned with the Accordingly, ICTRegie mobi- sector, the Authority wishes to design of an intelligent health- lizes stakeholders on both concentrate on the field of ICT & care environment which pro- the supply and demand side, the New Media, which includes motes the autonomy and self- encouraging them to join the developers of innovative determination of the patient. each other in thinking about content and the companies/ - The healthcare professional: opportunities for innovation. organizations which provide in- the themes Knowledge Man- It challenges them to innovative ICT solutions to dis- agement and Imaging diagnos- spire each other to arrive at seminate this content as televi- tics and image-directed treat- promising innovations using sion programmes, websites, ment are intended to provide ICT, encourages them to games and various multimedia the healthcare professional applications. with the level of knowledge seize the opportunities for and information required to innovation, and brings those Food and Flowers - The offer high-quality care to pa- opportunities together within ‘Food & ﬂowers’ sector involves tients. a national vision.” activities such as product en- - The healthcare manager: the hancement, horticulture, the theme ‘Tailor-made Care’ helps “Recognizing that real break food and confectionery industry the managers of healthcare -throughs come form cross and the raw materials supply institutes to cope with the disciplinary efforts, ICTRegie industry. The sector is incontro- complexity involved in provid- adopts a roadmap for dealing vertibly a strength of the Neth- ing customized care without with four areas of strength in erlands, largely due to a ﬁnely incurring additional costs. each of three disciplines. It balanced interplay between then sets forth procedures to technical control of production

Safety and security - Safety deed, ICT is itself ubiquitous and and increased productivity to and security involve applications is accepted as a ‘given’...... one which seeks to provide in the fields of defence, ICT has become a matter of added value to clients. Inte- peacekeeping, anti-terrorism course, a part of everyday life. grated, distributed, service- ori- This is a typically European con- operations, crime reduction and ented architectures, . . . . The cept in which the focus is on crisis management, as well as second reason is technological in people (from computer-centred ensuring the continuity of vital nature. The combination of huge computing to human-centred amenities such as energy, tele- computational capacity with vir- computing) and has formed the communications, public admini- tually unlimited networking and main theme of European ICT stration and logistics. storage capacity has enabled research for several years. It is today’s ICT to provide new also in keeping with the typical Education - All schools now forms of service rapidly and ac- European competences. have computers and broadband cessibly. Internet connections. . . . In the Product Software - or indus- next phase, state-of-the-art ICT try specific software - The COOK Report: All of the applications are to be integrated Authority intends to establish preceding are quotes from with developments in educa- how general knowledge, theories pages 18 -- 20 of the plan. tional reform. This will enable and (business) strategies from The common thread through new styles of teaching and learn- the world of software engineering to be undertaken, being not out is that the efforts are ap- ing can be rendered more widely only more efficient, but also plied to problems not in isola- applicable by means of product more effective with a greater tion, but rather from seeing software development. In doing degree of individualization, the problems as complex in- so, it is important to investigate opening up the vista of true ‘life- teractive systems. The which business strategies have long learning’. thread points out that knowl- proven successful in the interna- edge silos cannot thrive as Mobility - Mobility is con- tionalization of product software, silos any longer - that they cerned with both passenger and and what product and service freight transport using all avail- strategy has proven effective in are inevitably linked by the able modalities (car, bus, train, spreading the risks. growth of powerful comput- boat, aircraft) as well as the ers ubiquitous and unlimited relevant infrastructure. The ac- Enterprise Information storage and fast, capacious cessibility of major economic Systems- (EIS) support the networks. ICT is the common centres such as cities, industrial various business processes glue that binds it all together parks, harbours, and airports within an organization. Examples and that sustainable eco- remains a significant societal include Enterprise Resource nomic growth will be only problem. Predictability and the Planning (ERP) systems, data found through application of opportunity to choose between warehouses and workflow man- this glue. the various modalities based on agement systems. Companies in accurate information are impor- the financial services sector, tant aspects. . . . public administration and pro- In adopting a systemic way duction companies alike are now of looking at innovation, The ICT Sector - having to integrate their primary ICTRegie does something Ambient Intelligence - The processes with supporting ICT to that, as far as I know, has an ever greater extent. emergence of ubiquitous com- never before successfully at- puting, natural man-machine tempted. On page 21 it interfaces, wireless networks Services science - The first is states and ever more intelligent that the nature of ICT service (‘smart’) systems allows ICT to provision is changing. It is shift- “Innovation is a creative and be applied in creating an intelli- ing from an isolated activity dynamic process. Its essence is gent human environment. In- geared towards cost reduction the communication between re-

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 54 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 searchers, application developers In addition, the ICTRegie Advi- ess is one of throwing money and users which will lead to new sory Council will produce an an- at problems. ideas. Creativity is required to nual strategic report with its rec- generate those new ideas, but ommendations to promote focus. President-elect Obama has also to arrive at new combina- The report will be public and will pledged to create a national tions of existing ideas, the serve as the basis for debate. CTO. It seems to me that ‘cross-pollination’ between dif- The report, together with the ICTRegie should serve as a ferent ways of thinking, ‘outside recommendations of the Com- the box’ thinking and trans- munities of Interest will form the role model for the definition sectoral innovation. In many input for the National ICT inno- and function for this new and cases, this involves a multidisci- vation vision and ICTRegie’s an- very important position. plinary process. Success relies nual working plan.” p. 24 on leadership and vision.” . . . . Now since Kees Neggers COOK Report: What the pointed out to me that Ed The aspects of innovation may ICTRegie Plan seems to be Seidel is both a man who be represented as follows. attempting to institutionalize ‘gets it’ and, since assuming - Business opportunity: crea- is the cooperative, collabora- the directorship of the Office tivity in the formulation of re- tive nature of innovation on of Cyber Infrastructure at quirements and the concomitant the internet. Can we create NSF on September 1 2008 is ‘value proposition’. a framework for problem - Business model: creativity in in a position to do something the organization of the value solving where our best minds about our problems, I have chain in order to achieve the will solve problems in a way spent some time within the proposed added value for the that benefits Dutch society? Office of Cyber Infrastructure domain in question. portion of the NSF website - Technology: creativity in the The American looking at proposals, pro- manner in which the technology Approach Needs To grams, and papers. The 2003 is applied in fulﬁlling the re- Move in the Dutch Atkins report that established quirements of the client/user, the office is quite broad and with due regard for functionality, Direction sweeping. Exactly how we ergonomics, quality and ease of are to get from here to there I can’t help but notice and be implementation. is not nearly as well defined somewhat discouraged by the as it is in the ICTRegie five- contrasting American ap- “Mobilizing, inspiring and year plan. encouraging proach – an approach that has worked well when we had The 2007 OCI vision docu- ICTRegie has opted for a style of a capital base large enough ment – the most recent that I direction which facilitates rather to support it, but one that in than dictates. The Authority could find – is detailed from a the context of economic wishes to mobilize the parties on discipline point of view. The meltdown is likely no longer both the supply and demand process is more general and sustainable. Here are for the sides to take part in the Com- it seems to me that the proc- last few decades we have munities of Interest and to con- ess at this point is more im- tended to make large appro- sider opportunities for innovation portant. Publications on the priations to agencies like the together. It challenges them to OCI website seemed to re- inspire each other towards National Science Foundation flect a sizable cross-section of promising ICT innovations, and it and to measure progress by University research interests encourages them to seize and the percentage with which we without being tied down in elaborate the opportunities thus can increase those appropria- any way that reflects with identified. tions every year. The proc- any precision how they will

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 55 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 contribute to economic But these tools involve far the Gigaport project today. growth and development in more than just Lightpaths, (December 31, 2008.) the United States. they involve grid software, Web 2.0 services and highly In short the COOK Report While this is in part explain- trained technical people who concludes that the technology able by the mission of the are only now beginning to works and that the most sig- National Science Foundation develop collaborative re- nificant variable will be the to support basic research, search programs in interdis- speed of dissemination. The one must wonder whether a ciplinary fields like those the speed and resulting benefits mission in basic research iso- Dutch have pioneered at their to national economies will be lated from an understanding GigaPort over the last five determined by the ability of of its impact on national eco- years. governments to invest in na- nomic productivity in our tional fiber networks as infra- country will be sustainable in What I saw at the Supercom- structure that extends the the context of our recent puter meeting in Austin holds benefits of ICT primarily to losses of several trillions of huge promise and is very ex- the society at large rather dollars as a result of the un- citing. The booths were filled than to carrier shareholders. regulated Wall Street binge? with what would look like magic, to the non-technical If I were President Obama Nevertheless based on a brief person. But those who create what then would I do? meeting with Ed at SC08 and the magic realize that they on discussions with others I must begin to develop a pro- 1. Figure out how to acquire conclude that he understands gram to transfer what they for cash or for other “sua- the issues discussed here do to the commercial and sion” a federal IRU on fiber should be given enthusiastic educational world. pairs on Level 3's, ATT and support as he undertakes this Verizon’s national foot- important new mission. The Dutch have been working prints. on a small scale with Phillips, 2. Give federal dispensation Some Thoughts on IBM and Unilever for several for the RONs (Regional What is Needed years. They understand what Optical Networks) to join has to be done. Managing their fiber into those pairs In my opinion SURFnet is the these tools to facilitate tech- forming a National Innova- leading optical network in the nology transfer is a complex tion Network. world and the SURFnet build- task that requires deliberate 3. Offer encouragement for ers have found that while and careful planning. enterprises to participate - they have created many use- long term - in e-science on But my experience also has ful tools, researchers and en- this innovation network shown that these tools are terprise scientists in Holland with an understanding that real, extremely powerful and do not use these tools in their nothing miraculous will very much deserving of care- everyday work and, at a happen over night. ful dissemination. After all, minimum, need training in 4. Elevate the role the NSF the interviews herein show understanding how the tools Office of Cyber infrastruc- how SURFnet has, at least can be a productive use of ture to a much more public since iGrid2002, been trying their time. and permanent focus as to very carefully nurture the the focal point for enter- process that has just come to prise outreach education completion with the end of coordination.

5. The FCC, the new National modeling on a large scale can “The focus on video as the CTO, and the directors of be invested into the building motivation for true broad- Obama’s infrastructure of our human capital. band [must be] temporary.” implementation team should let all the states A knowledgeable friend com- “Network applications involv- know that fiber intercon- plained that requiring any ing access to, and sharing of nection with their nearest national fiber network to un- large volumes of binary data RON is strongly encour- bundle one or more fiber as the basis of information, aged. pairs was tantamount to forc- and ultimately as a basis of 6. There should be some sig- ing it to commit suicide and knowledge, are highly devel- nificant education efforts to causing the implosion of oped, but are not so visible in undertaken by the admini- the telecommunications in- the world of entertainment stration’s media people - dustry, putting additional and social networking, as via and similar internet huge numbers of out of work. they are in the realm of re- tools YouTube - to explain Faced with this concern, I search. to Americans why this is must ask whether we subsi- worth while. This “stuff” is dized the horse and buggy But soon corporations will intensely Visual and if you whip industry at the expense learn to follow in the foot- have had no immersion in of cars because we were steps of the research it, the impact is easy to afraid of unemployment? community to handle and dismiss. benefit from the knowl- 7. The cost of doing this is I submit that this opinion is edge implicit in such data- not HUGE, the payback exaggerated and assert be- sets, whether for health- immense. after the crash sides that we cannot afford care or for other business of the ponzi economy it not to wisely invest what processes, or for new would be nice to have economic resources we forms of education, that some vision to inspire stu- have left. complement web-page and dents that there was video (more traditional) ‘con- something more meaning- Be it a new ARPA or an tent’.” ful than Wall Street in American ICTRegie our their future. (See new administration MUST “Even in the days when walls Paul Krugman, The Madoff Econ- grasp the present oppor- of your home are live dis- o m y ” tunity to sweep out the plays (the walls themselves, http://www.nytimes.com/2008/1 wreckage of cronyism and as extensions of current 2/19/opinion/19krugman.html?sc p=1&sq=krugman%20ponzi&st= speculative capital. In the OLED developments, not just cse ) national and public interest screens), it will be the knowl- like Roosevelt did in the edge behind the images, and These are highways of light 1930s, it must build out a the ways they are used to that will subsume all forms of “rural electrification system” inform and educate, as well communication education, - this time an interstate as entertain, that will matter commerce and knowledge highway system of optical most.” transmission. The sooner fiber. they become ubiquitous, the And just before Christmas more broadly their educa- I end with a reminder of Har- Harvey sent me a private tional and collaborative ef- vey Newman’s quote of a note which he has given me forts in all forms of science - month ago. permission to publish. I was especially those requiring

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 57 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 pointing him to JP Rangas- business model of the Inter- us out of the past. wamis’ blog saying net ?" And if we do not do this - This man gets it!! He is In this case what they are there is strategic as well as pushing BT toward becoming asking is: How do companies economic risk. Since any na- a services based cloud com- ("how do I") make a profit tion can do this.” puting platform. out of this ? I will be further exploring the Harvey Newman: I agree So I say - implications of this via an in- with this. And it makes me terview with Harvey in the think - Often, to make one'sn Only the highway-building March issue. in-cumbent, one is con- model ("building the nation strained to say "What is the from the bottom up") will get

To obtain the entire report: http://www.ictregie.nl/publicaties/nl_08-NROI-258_Advies_ICT_infrastructuur_vdef.pdf

Introduction pp. 1-3 strengthening and appropri- Working with an all optical ate dynamism of a Dutch ICT network made the assign- knowledge infrastructure ment of light paths possible, This issue examines the con- geared towards high social first to institutions, then to tinued evolution of both fiber and economic yield.” departments and ultimately infrastructure and optical to individuals. networking in research, education, and tech transfer A New National The five year GigaPort project networks and network pro- Infrastructure, running from January 1 2004 jects in the Netherlands. It pp. 4-14 through December 31 2008 also looks at the Dutch plan- made optical private net- ning process for directing Kees Neggers, SURFnet Man- works possible for SURFnet economic investment in ICT aging Director, explains how members. It also made pos- in ways designed to achieve, innovative technologists used sible the establishment of a through innovation, maxi- the intersection of smart resource management plat- mum economic impact. Dutch telecom liberalization form built on a foundation of policies, and then the dot grid computing, a layer of The introduction and conclu- com boom to build competing web services allowing users sion extensively discusses the fiber rings around Amsterdam to connect to applications and programs of ICTRegie – the creating an attractive envi- to instruments as well as su- Dutch ICT Research and In- ronment that allowed the city per computers. On top of the novation Authority. They re- to become a global hub for resource management plat- flect a process set up in 2005 layer three traffic at AMS-IX form is one of generic e- whereby Government re- and layer one and two optical science services (the light search funding is put through light waves at NetherLight. blue box in the diagram on a process that is designed to page 11) in other words create proposals that are in Kees explains the sequence software tools that support keeping with nationally de- of events leading to SURF- research in individual fields. fined goals allied to the na- net’s solving the last mile tion’s social needs as well as connection by means of fiber Within GigaPort one has cut- its economic strengths. for all its members in 2004 ting edge technology sup- and becoming the first na- porting not only basic re- The mission statement is tional all optical research search but also tech-transfer what President Obama should network at the end of 2005. from universities to compa- emulate in his Office of the This enabled what they call nies like Philips, Unilever and CTO. “First to introduce unity hybrid networking – doing as IBM. It also has valuable ex- and consistency to the stra- many tasks a possible at perience with the complexi- tegic direction of the ICT re- layer one and two and as few ties of making it work search and innovation by tasks as possible with much through ways to build and means of the development of more expensive electrical and administer the corporate a national strategy that en- routed layer three networks. networks, to the software joys broad support and, sec- tools needed to take advan- ond, to ensure ongoing tage of the lightpaths to

© 2009 COOK network consultants 431 Greenway Ave. Ewing, NJ PAGE 61 THE COOK REPORT ON INTERNET PROTOCOL FEBRUARY 2009 needed training to enable us- Hybrid Networks and most optimal way to achieve ers to benefit from the tech- e-Science Future what you want to have done. nology at their disposal. Development, pp. 15 What all this means – and Most significant is the deci- -40 this is the most funda- sion of ICTRegie to recom- mental thing to get one’s mend continued support for Cees de Laat takes us mind around – is that your SURFnet and GigaPort as a through the events that be- network becomes just basic platform for technology gan with SURFnet’s 2001 depart of your programming development in the Nether- cision to work with Starlight environment. lands. As Harvey Newman in Chicago and CENIC in San says at the very end of this Diego as well as a few other Normally you have your data issue, while the private sector locations to explore the use and your computing and a is critical without roads open of dedicated light waves of “stupid network” that is an to all we will be placing a one to ten gigabits. The ex- unmodifiable “given” that you huge handicap on our possi- periences of the iGrid2002 have to play along with. But bility for further progress. meeting led Cees and his col- here your network is just leagues to begin work with part of your toolset.” As Kees Neggers said: In our circuit switched layer 2 light- view innovation needs the paths wherever possible. Later he elaborates: “In network and hardware re- our group back home we are sources plus the software Working with extremely de- creating objects for virtualiz- and the people that have manding users who needed ing and programming wave- the knowledge of how all their own lightpaths led Cees length switches and photonic of these layers interact and his group to think about devices which directly talk to with each other and in ad- tailoring the network to the fiber; Micro-Electro- dition to this knowledge needs of the user so that, Mechanical Systems (MEMS) you still need a major out- unlike layer 3 routed net- devices that can connect fi- reach effort to involve the works that are stupid pro- bers so that we have flexibil- users for which it has crustean beds into which eve- ity at the fiber layer. And at been created. rything must fit, the networks the Ethernet layer, we can do could become much more similar things manipulating [And Cees de Laat added]: flexible and adaptive for the Virtual Local Area Networks From the light blue box population that the network (VLANs). We address the you need people working is designed to serve. Ethernet layer and the packet with every discipline to routing when we need that. make it happen, because As Cees explains: “what I am the people in the disci- describing is a kind of pecu- If you can manipulate all plines themselves cannot liar programming language or these layers and have also do that. You can not ex- programming environment vertical and horizontal pect the biology professor where your networks are just knowledge in every layer, to understand how to go subroutines, your data are you can do the magic and out and identify and hire more subroutines, your solu- you get a perfectly inte- the people necessary to tions to solve the problem grated multilayer hybrid teach him and his col- are also subroutines. You network that is optimized leagues how to use the then say optimize and solve for your application.” high performance tools. and it will then work out the

Cees goes on to elaborate the Interactive Networks, At the end he concludes: kinds of e-Science projects p. 41-49 With the information that these lightpath networks given by the network, and support. He then demon- currently we support con- I have written up with illus- strates in some detail his VL- tinuous measurement of trative photos a demonstra- e (e-Science Virtual Labora- delay, jitter, bandwidth tion by Rudolf Strijkers of tory) applications. and throughput, one or what a user interface to ap- more operators at once plication embedded networks can write programs that I said to him: what you are looks like and the capabilities automate decision- explaining is how to build a it permits. networking environment making. This opens the that combines this use of way for automated net- What we discuss is a proto- equipment, technology, work adaptation in a user- type of a user interface with and energy in such a way friendly environment. a multi-touch-sensitive to give the user applica- screen that allows a user to tions that meet his de- For example, now we can tap on the tools he wishes to mands at the minimal say, certain paths should select and with his finger to avoid busy parts of the necessary cost? draw the paths he wishes in network. By using only order to activate nodes on a standard functions in De Laat: “Yes. But to do programmable network. this one must expose the Mathematica, it is already network to the users so possible to write a simple As Rudolf pointed out: This is program that uses the that they understand what the first prototype of what we real-time throughput in- is happening and can see are calling “Interactive Net- formation to continuously the benefits of not always works”. In interactive net- reroute one or more paths working at the highest works humans become an and most expensive layer that avoid busy parts of integral part of the control the network. -- namely the routed system to manage the next- layer. One needs to make generation of programmable sure that your users un- networks and Grids. The derstand the benefits of main design principle is this; Conclusion pp. 50-58 operating in different lay- by virtualizing the configur- ers of their optical net- able and programmable The technology that is dis- work. The users need to properties of network ele- cussed here is inevitable and grasp that if they endow ments as software objects, inexorably extensible and us- their applications with the any aspect of a network in- able at reasonable cost where ability to intelligently frastructure can be manipu- open access fiber is available. traverse the layers they lated from computer pro- The countries that implement can open up all manner of grams. What we show here is this openly and broadly as an increased possibilities to an implementation of an in- infrastructural system of better performance at less teractive control system con- roads and highways will de- cost.” cept for user programmable rive enormous competitive networks, which applies the educational and economic For the most part the rest of architectural concepts we benefit. Those that don’t will interview describes how rout- have developed in our re- never see either the educa- ing or path finding is done in search. his hybrid optical network.

tional or the economic bene- wreckage of cronyism and fits. speculative capital. In the asking is: How do companies national and public interest ("how do I") make a profit We touch on the OptiPuter like Roosevelt did in the out of this ? and OptiPortals and their sig- 1930s, it must build out a nificance for science educa- “rural electrification system” So I say - Only the highway- tion. We also extensively dis- - this time an interstate building model ("building the cuss the Netherlands ICTRe- highway system of optical nation from the bottom up") gie five year plan and show fiber. will get us out of the past. how it can be used to further distributed edge based col- I conclude with a mention of And if we do not do this - laboration and consensus in BT becoming a services there is strategic as well as allocating economic re- based cloud computing plat- economic risk. Since any na- sources. We offer a seven form that I made to Harvey tion can do this.” point national fiber infrastruc- Newman. ture plan for the new admini- COOK Report: I will be fur- stration. Harvey Newman: I agree ther exploring the implica- with this. And it makes me tions of this via an interview Be it a new ARPA or an think - Often, to make one's with Harvey in either the American ICTRegie, our case for innovating at an in- March or April issue. new administration MUST cumbent, one is constrained grasp the present oppor- to say "What is the business tunity to sweep out the model of the Internet ?"

In this case what they are

A Note from the Editor on the February 2009 Format and Presentation This issue has an Introduction; an interview with Kees Neggers; another with Cees de Laat; a presentation by Rudolf Strijkers and a closing essay. The Symposium discussion is postponed for a month

Text, URLs and Executive Summary: I have attempted to identify especially noteworthy text by means of boldface for REALLY good “stuff” . Also the proper Executive Summary in this issue continues. I hope you find it useful. Feedback welcomed. You will also find live URL links and page links in this issue.. (I am also no longer changing Brit- ish spellings of things like fibre to the American fiber. )

Thanks to Sara Wedeman - see sarasworld.blogspot.com/behavioraleconomics/ for assistance with the masthead logo. Captain Cook now charts direction by looking at a compass rosette.

Coming in the March 2009 issue - out about January 31 probably a discussion with Frank Coluccio on ﬁber to the desk top and taking copper out of the networks. On December 29th I completed a marathon 4 plus hour interview with Harvey Newman. The March issue will deﬁnitely have symposium discussion - I am not sure at this point how i will handle the Newman and Coluccio material

I am omitting the contributorsʼ page since a cumulative list may now be found at http://www.cookreport.com/index.php?option=com_content&view=article&id=121&Itemid=74

Contents

ICT and E-Science as an Innovation Platform in The Netherlands A National Research and Innovation Network What Can the US Learn from Dutch Experience?

Before Resource Use Careful Deliberation and Coordinated Planning Needed p. 1 E-Science p. 2 Adopting a New National Infrastructure How Fiber Became the Basic Platform for Education, Commerce, Technology Transfer, and Research in the Netherlands

Sparking the Fiber Revolution p. 4 The Signiﬁcance of AMS-IX p. 5 The Germination of the Hybrid Network Approach in 2001 p. 6 Adoption of Lightpaths - Not only for Science but also for ICT Departments p. 7 Dutch to American Analogies and the E-Science Virtual Laboratory p. 9 The Optical Network and its Platform for Collaboration Become Basic Infrastructure p. 11 The Basis for a Future Internet? Optical Hybrid Networks and e-Science as Platforms for Innovation and Tech Transfer p. 15

How the World Changes When You Give the User his Own Network p. 17 Treating the GigaPort as Tech Transfer Infrastructure p. 20 Lambda Grid Developments and the Need for Hybrid Networking p. 21

WiVR: a Window into Virtual Reality p. 22 The CineGrid Project p. 23 What These Large Scale Projects Teach Us About User Populations p. 24 How Low Can You Go? p. 26 The Global Fiber Infrastructure p. 26 The Signiﬁcance of SURFnet6 as a Photonic Network p. 29 A Photonic Switch p. 30 Resource Description Framework Language p. 32 Network Description Language as Routing Protocol p. 35 Terabit Networking p. 38 Interactive Networks A User Interface for Application Owned Lightpath Networks p. 41 Running Mathematica p. 47

Where Is All This Headed? p. 50

For Further Investigation - the OptiPuter p. 51 Itʼs the Process and the Planning p. 52 The American Approach Needs To Move in the Dutch Direction p. 55 Some Thoughts on What is Needed p. 56

Appendix: E-Science Recommendations p. 59

Executive Summary p. 61

What is the COOK Report?

For more information visit http://www.cookreport.com/

This book is produced and distributed by and under license with Cook Network Consultants tHe C oo k re p ort on internet p rotocolBuilding a

BuildingBuilding AA nnaationtionaall KnowledgeKnowledge IInfrnfraastructurestructure

HowHow DDutcutchh PrPraagmgmaatismtism nurturesnurtures aa 2121stst CCenturyentury EEconomyconomy n a tion a l Knowledge I nfr a structure

The COOK Report ISBN 978-90-814102-2-9 On Internet Protocol