2-3 Trends on Network Testbeds in the World

MURASE Ichiro

becoming to the infrastructure of any kinds of R&D. in U.S. includes various strate- gic networks and projects. And in Europe there are collaboration among many network testbeds leaded by EU. Regarding characteristics on network testbeds, we have three issues. One is an adoption of photonic technology. Two is a focus on middle ware technology. Three is an attention on applica- tion technology. In the future, each network testbeds will have an originality, and they have many collaboration each other over the border. Keywords JGNⅡ, , Network testbed, Next generation network, Photonic network

1 Introduction the network. (3) New network technologies are implement- The Japan Gigabit Network (JGN) began ed on the network. operations in 1999, and was hailed at the time However, the above three conditions are as the fastest network testbed in the world. Its not necessarily absolute; a network may be successor, the JGNⅡ, has been in operation regarded as a network testbed even if it fails to since 2004. Network testbeds are now operat- meet one of the conditions. It is widely known ing in North America, Europe, Asia, and else- that the JGNⅡ network is currently hosting a where, with the incorporation of photonic wide range of R&D, with numerous associated technology, middleware research and develop- research institutes linked to this network. The ment, and the investigation of a range of appli- network also incorporates the latest photonic cations. This paper provides an introduction to technology. these current trends in R&D of network test- Network testbeds are designed to provide beds. structures simulating those we assume will be required for future networks. These assump- 2 What are network testbeds? tions naturally give rise to the above criteria stipulating the general conditions that must be Network testbeds are networks designed satisfied by a testbed. As an additional mainly for the development of network tech- requirement, to encourage R&D, the AUPs nology. (In this paper, the terms “testbed net- (Acceptable User Policies) for these testbeds works” and “network testbeds” are inter- must specify restrictions on commercial traffic changeable.) The discussion in this paper and clarify that no SLA (Service Level Agree- assumes that network testbeds satisfy the fol- ment) is provided with use. lowing criteria. (1) R&D takes place on the network. (2) Multiple research institutions are linked to

MURASE Ichiro 21 3 General trends in network test- of grid technology, for example). beds The following sections will describe the major network testbeds in North America, Existing testbeds have often been designed Europe, and Asia, with the above observed with a focus on ultra-high-speed communica- trends in mind. tion. However, given a worldwide trend toward lower communication costs, it is 4 Network testbeds in north amer- becoming increasingly evident that universi- ica ties and research institutes are not prioritizing the provision of ultra-high-speed network test- 4.1 CA*net4［1］ beds. Thus, testbed design is presently facing CA*net4 is ’s coast-to-coast pho- a dramatic shift in strategy. tonic network developed for research and edu- The characteristics of current network test- cation. It connects all regional networks of beds can be described through the following Canada and supports the use of advanced observations. applications. (1) We are witnessing an emphasis not only on Expanding on the achievements of its pre- the network testbed itself, but also on the decessor, the CA*net3 project (involving the R&D projects that the testbed can support. construction and technological evaluation of a (2) R&D projects normally involve multiple photonic network optimized for IP communi- network testbeds, and so the relationship cations), the CA*net4 project has as its main between network testbeds and R&D pro- technological target the implementation of jects is becoming increasingly complex. UCLP (User-Controlled Light Path). The (3) The spectrum of targets in R&D projects is UCLP is a user-controlled network architec- extending to middleware as well as to ture that allows end users to manage and con- fields in science and technology, with even trol their assigned optic fibers and bandwidth further expansion in scope anticipated. at the GigaPoP nodes. (4) Energetic R&D is now underway for a The CA*net4 network topology is shown variety of middleware (involving the use below.

Fig.1 CA*net4 network topology (Excerpt from http://www.canarie.ca/canet4/library/c4design/large_file_transfer.ppt)

22 Journal of the National Institute of Information and Communications Technology Vol.52 Nos.3/4 2005 4.2 Abilene (Internet2)［2］ Abilene is a research testbed network in the US dedicated to Internet2 projects, and is designed for the development of advanced network technologies, innovative Internet applications, and commercial Internet imple- mentation. Abilene use is presently limited to education, e-learning, research, and clinical studies among institutions of higher education, and it is not open to commercial traffic. Access is limited to Internet2 members and Fig.2 topology these educational institutions. (Excerpt from Abilene is operated by the Internet2 Con- http://abilene.internet2.edu/images/abilene- current.gif) sortium, a group formed by the University Corporation for Advanced Internet Develop- ment (UCAID) and a number of corporate (National Science Foundation). partners. Figure 3 shows a summary of the major Since it is not directly associated with network testbeds and associated projects national governmental policies, Abilene established primarily by the NSF. receives no special funding from the federal Table 1 provides details of the network government, and is solely supported by the testbeds and the related projects shown in Fig. Internet2 membership fees and connection 3. charges collected from participating institutes. HOPI is an Internet2 project to implement However, research using the network is in part a hybrid shared-IP-packet and switched-opti- funded by grants based on various govern- cal-wavelength network, with the deployment mental policies and initiatives such as the of testbed facilities planned for 2006. The NGI. HOPI testbed is being constructed using the The operational costs are covered by capi- Abilene, RON, and NLR networks. tal investment by universities, matching The Regional Optical Networks form a investments by businesses, and membership photonic network testbed operated by Fiber fees (totaling $3.3 million/year). The connec- Co. founded by the Internet2 to support tion charges for OC-192 POS/10GigE total regional photonic fiber network initiatives $490,000/year. University members of Inter- aimed at research and educational institutions. net2 contribute capital investment totaling $80 The MAN LAN is a high-performance million annually for the expansion of universi- exchange point located in City ty facilities, network connection fees, and designed to facilitate interconnection among application development. Corporate members US and international research and education contribute in a variety of ways, from monetary networks. It was built through a collaborative support to the provision of network-related effort between the Internet2, NYSERNet, and equipment, in addition to payment of the Indiana University. Connection charges are required membership fees. collected on a monthly basis. The NLR is an ultra-high-speed network 4.3 Range of network testbeds in the owned and operated by researchers. The Inter- US net2 provided a total of $10 million in funds to Abilene (Internet2) is the best-known net- cover a five-year period, and has access rights work testbed in the US, but there are a multi- to the 10-Gbps circuit. The Internet2 organiza- tude of additional network testbeds and asso- tion is using the NLR along with Abilene as ciated projects, mainly established by the NSF foundations for the development of a hybrid

MURASE Ichiro 23 Fig.3 Major network testbeds and associated projects mainly established by the NSF (Drawn from references provided by Mitsubishi Research Institute, Inc.)

Table 1 Major network testbeds and associated projects by the NSF

24 Journal of the National Institute of Information and Communications Technology Vol.52 Nos.3/4 2005 optical packet environment. context of an optical wavelength experiment The Teragrid offers 20 teraflops in com- conducted with MAN LAN. puting capacity using computer resources dis- tributed among five sites. The NSF provided 5 Network testbeds in Europe $53 million in funding for four of these sites in 2001. The NSF Middleware Initiative effort In Europe, the EU (European Union) is now involves research and development taking the initiative in establishing the so- involving the Internet2, EDUCAUSE, and called “e-Infrastructure”. This infrastructure SURA. will promote the coordinated use of research The “I2 Middleware Initiative” is a project and information resources scattered around to provide core middleware services to Inter- Europe (involving GRID calculation, high- net2 member universities. speed networks, storage, etc.) to create a The ARENA project involves the creation framework for the development of e-science. of a database of information on advanced The European network testbed—such as research and education networks around the GEANT, EGEE (a GRID research project), world. and DEISA—all hope to benefit from the syn- This is a joint project between researchers ergy arising from the e-Infrastructure. from universities and elsewhere to rebuild existing Internet architecture to provide 100- 5.1 GEANT and GEANT2［3］ Mbps access rates to homes and small busi- GEANT is a network testbed promoted by nesses. This project has been awarded a five- DANTE (Delivery of Advanced Network year grant from the NSF Information Technol- Technology to Europe), with interconnection ogy Research (ITR) Project. The Internet2 among the NRENs (National Research & Edu- organization provides technical support and cation Network) of different European coun- two principal researchers for this project. tries. The purpose of the testbed is to provide The Pacific Wave project is part of joint a data-communication environment beyond efforts between CENIC and PNWGP, and has the gigabit level. GEANT was in operation as its goal the construction of a network con- from Nov. 2000 to Oct. 2004, and GEANT2 nection facility covering the entire US Pacific began operations in Sept. 2004. The number coast. This will promote efficiency in IP traf- of target NRENs for interconnection is fic by relaying such traffic from other net- assumed to be 30 (See Fig. 1). The band- works, as well as through greater overall effi- widths of the GEANT and GEANT2 back- ciency in network transmission. bone circuits are 10 Gbps and 40 Gbps (10 PlanetLab is an open, decentralized plat- Gpbs×4), respectively; there appearing to be form for developing, deploying, and accessing strong demand for the establishment of faster, worldwide network services. The Abilene ultra-high-speed backbones. observatory (a project to support the collection and distribution of Abilene network data) pro- 5.2 SURFNET［4］ vides the eight core router nodes for Planet- SURFNET is a national research network Lab. in Holland that links research and educational The GLIF is a worldwide Lambda-based institutions such as universities in the country research facility for the development of appli- through high-speed networks. It was devel- cations and middleware on the LambdaGrid. oped as a network platform for the GigaPort This facility was founded by researchers at Project (involving the Dutch government, pri- SURFnet, the University of Amsterdam, and vate businesses, and educational and research NORDUnet at the 2003 Annual Global institutions), and is now being applied to R&D LambdaGrid Workshop. The Internet2 organi- of applications relating to electronic collabora- zation is participating in this project in the tion, e-business, and more. As of 2005, the

MURASE Ichiro 25 Fig.4 GEANT2 network topology (Excerpt from http://www.geant2.net/upload/pdf/GEANT2_brochure_2nd_edition_final.pdf) backbone speed of SURFNET5 is 30 Gbps, but there are plans to increase this to 80 Gbps. Figure 5 shows the network topology of SURFNET5.

5.3 EGEE［5］ EGEE is a collaborative effort among 27 countries in Europe to develop grid technolo- gy and grid-based services, and is based on the GEANT network. The goal of this initiative is to provide an environment in which researchers may access computer resources regardless of physical location. The main objectives of this initiative are as follows. (1) To construct a stable grid network and increase access to computer resources (2) Development and maintenance of middle- ware for more reliable service Fig.5 SURFNET5 network topology (3) To attract researchers from industry and (Excerpt from http://www.surfnet.nl/en/organi- sation/network/flasheng/map.html) academia and to provide opportunities for advanced support and training Below are some examples of actual EGEE Uses the grid foundation to store and ana- applications. lyze petabyte-level data obtained through • Large Hadron Collider Computing Grid actual experiments and simulations of the (LCG) high-energy physics experiments performed at

26 Journal of the National Institute of Information and Communications Technology Vol.52 Nos.3/4 2005 CERN • Biomedical Data mining of petabyte-level data within the genome database, indexing of medical databases, and on-demand computing of het- erogeneous data

6 Network testbeds in Asia

6.1 CERNET2［6］ In China, efforts to create a next-genera- tion Internet on a national scale began official- ly in Aug. 2003, with the following three Fig.6 CERNET2 network topology goals. (Excerpt from • Creation of an advanced network to support http://www.ampath.fiu.edu/CANS2004/nov%2 future research 030/Jianping%20Wu-%20nov30.ppt) • Creation of a test environment for new tech- nology and new business applications the network has shifted to providing an • Cultivation of an industry of IPv6-compati- advanced research environment for next-gen- ble products eration technologies based on QoS, Multicast, CNGI consists of three layers—(1) a layer IPv6, and MPLS. that establishes connection with the interna- The Ministry of Information and Commu- tional NGI; (2) multiple backbone networks, nication (MIC) of Korea presents the policies metro networks, and the GigaPoP system that and the project goals of KOREN based on connects them; and (3) a multitude of campus overall public policy, and formulates the cor- networks. The CERNET2 network is different responding basic plans. from CERNET, and it is intended to serve as The National Computerization Agency the largest backbone network for China’s (NCA) of Korea is in charge of the operation next-generation Internet, the CNGI. The net- of the network, organization and management work was planned as part of a national NGI in of development meetings and user forums, and 2001; trial operations began in 2003. It has support for international research collabora- provided IPv6 services since April of 2004. tion. The KT Corporation is in charge of the The network topology of CERNET2 is pre- construction and modification of the network, sented below. operation and maintenance, and support and management of member institutions. 6.2 KOREN［7］ The KOREN network topology is illustrat- KOREN is a non-profit research network ed below. provided to universities, research institutions, and industry research organizations. It offers a 6.3 SingAREN［8］ research environment for the development of SingAREN is a research and education high-speed communication devices and appli- network in Singapore, with a main focus on cation services, and is positioned to support joint international research and broadband R&D of basic technology and associated application development. The network was applications, and more broadly, Korea’s IT created as a national project to reinforce the vision of an Information Superhighway. activities of the research and education com- Between 1995 and 1997, KOREN mainly munity in Singapore and to promote interna- supported the development of network tech- tional cooperation, and began operations in nology and associated devices, but since 1998 2001. Universities and research institutions

MURASE Ichiro 27 Fig.7 KOREN network topology (Excerpt from http://www.ccirn.org/general/AP%20Net- works%20at%20CCIRN.ppt) are participating in R&D of grid computing, photonic networking, and broadband applica- tions such as those involving IPv6. A total of five national universities and Fig.8 SingAREN network topology research institutions are involved in the opera- (Excerpt from http://www.singaren.net.sg/con- nect/netmap.shtml) tion and maintenance of SingAREN. Among these entities, the National University of Sin- gapore (NUS) plays the central management among university researchers. Today’s ThaiS- role and performs duties outside of normal arn3 represents the third generation of this net- operations, such as management of the execu- work, and was constructed based on the Thai tive office. government’s IT2000 Plan (originally sched- In Oct. 2003, SingAREN was transformed uled for implementation from 1996–2000, but into a non-profit organization and currently the start date was delayed until 2002 due to receives no funding from the Singapore gov- the country’s economic crisis). ernment. The topology of the ThaiSarn network is The network topology of SingAREN is illustrated below. shown below. 7 Future trends in network test- 6.4 ThaiSarn［9］ beds ThaiSarn is a network for research and higher education, connecting researchers in Network testbeds are no longer limited to Thailand with their counterparts throughout network technology research; specifically, we the world. The major nodes within Thailand are entering a period during which R&D will are linked by 155–620 Mbps ATM circuits. expand into new fields—from grid and IPv6 The current goals of network operations middleware to general applications in science include research on application technologies and technology. In the future, the unique fea- in remote education, remote medicine, TV tures of the individual network testbeds will conferences, VOD (video on demand), and the no doubt come into play when the internation- practical implementation of these technolo- al testbeds are connected both physically and gies. through joint research. At the same time, the ThaiSarn operations began in 1992 with construction and operation of network testbeds the aim of promoting Internet use, especially throughout the world will most likely continue

28 Journal of the National Institute of Information and Communications Technology Vol.52 Nos.3/4 2005 Fig.9 ThaiSarn network topology (Excerpt from http://thaisarn.nectec.or.th/htmlweb/image/ThaiSarn-Map-030320052.gif) to proceed hand-in-hand with individual Acknowledgements national ICT strategies. In Japan, as well, the key to future net- I am deeply grateful to Mr. Masaki Ishig- work testbed design will lie in our vision of a uro, Mr. Shingo Inoue, and Ken-ichiro Akai future information society. (all at Mitsubishi Research Institute, Inc.) for the valuable information they provided in the course of preparation of this paper.

References 01 http://www.canarie.ca/about/index.html 02 http://www.internet2.edu/ 03 http://www.dante.net/ 04 http://www.surfnet.nl/ 05 http://public.eu-egee.org/ 06 http://www.edu.cn/HomePage/cernet_fu_wu/internet_2/index.shtml 07 http://www.koren21.net/eng/index.php 08 http://www.singaren.net.sg/start.php 09 http://thaisarn.nectec.or.th/htmlweb/index.php

MURASE Ichiro 29 MURASE Ichiro Mitsubishi Research Institute, Inc. Information Security

30 Journal of the National Institute of Information and Communications Technology Vol.52 Nos.3/4 2005