DOCSLIB.ORG

Special Guest at D-ITET: Nick Mckeown

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Special Guest at D-ITET: Nick Mckeown Special Guest at D-ITET: Nick McKeown Professor of Electrical Engineering and Computer Science at Stanford University Faculty Director of the Open Networking Research Center Software-Defined Networks and the Maturing of the Internet Monday, 24 November 2014, 15:15 Auditorium ETF E1, Sternwartstrasse 7, 8092 Zurich Department of Information Technology and Electrical Engineering Abstract The genius of the pioneers of the Internet was to keep the network of links and routers – the “plumbing” – dumb and minimal, placing as much of the intelligence as possible in the computers at the edge. Our computers at the edges could be upgraded over time to add new features – such as congestion control and security – without having to change the network. A streamlined network could focus on forwarding packets as fast as possible. A simple network with distributed control allowed for organic, explosive growth in the 1990s, with small businesses popping up everywhere to offer Internet service. But over time the network became more and more bloated, straying far from the original intent, with thousands of complicated features locked inside closed, vertically integrated routers. Networks became harder to manage, and those who own large networks fell under a stranglehold from their equipment vendors. Innovation was slow, equipment was unreliable and profit margins were through the roof. The networking industry of the 2000s turned into the mainframe industry of the 1980s. Along came compa- nies building data centers with thousands of switches and routers, with a pressing need to place the network under their control. Over-priced firewalls and load-balancers were replaced with homegrown software running on servers. Routers and switches were simplified, making them more reliable, lower- cost and lower-power. The entire network was placed under the control of software created using modern software practices. The software defined network (SDN) was born. In turn, SDN made it easier to use the servers, storage and network more efficiently by virtualizing the network. In this talk Nick McKeown will explain why – despite all the hype – software-defined networks were really just an inevitable next stage in the maturation of the Internet. And he will explain what he thinks happens next. Biography Nick McKeown (PhD/MS UC Berkeley ’95/’92; B.E Univ. of Leeds, ’86) is the Kleiner Perkins, Mayfield and Sequoia Professor of Electrical Engineering and Computer Science at Stanford University, and Faculty Director of the Open Networking Research Center. From 1986-1989 he worked for Hewlett- Packard Labs in Bristol, England. In 1995, he helped architect Cisco’s GSR 12000 router. In 1997, Nick co-founded Abrizio Inc. (acquired by PMC-Sierra), where he was CTO. He was co-founder and CEO of Nemo (“Network Memory”), which is now part of Cisco. In 2007, he co-founded Nicira (acquired by VMware) with Martin Casado and Scott Shenker. In 2011, he co-founded the Open Networking Founda- tion (ONF) with Scott Shenker. Nick is a member of the US National Academy of Engineering (NAE), a Fellow of the Royal Academy of Engineering (UK), Fellow of the IEEE and the ACM. In 2005, he was awarded the British Computer Society Lovelace Medal, in 2009 the IEEE Kobayashi Computer and Communications Award and in 2012 the ACM Sigcomm Lifetime Achievement Award. He received the IEEE Rice Award for the best paper in communications theory. Nick’s current research interests include software defined networks (SDN), network verification, video streaming, how to enable more rapid improvements to the Internet infrastructure, and tools and platforms for networking research and teaching..
Load more

Recommended publications
  • Openflow: Enabling Innovation in Campus Networks
    OpenFlow: Enabling Innovation in Campus Networks Nick McKeown Tom Anderson Hari Balakrishnan Stanford University University of Washington MIT Guru Parulkar Larry Peterson Jennifer Rexford Stanford University Princeton University Princeton University Scott Shenker Jonathan Turner University of California, Washington University in Berkeley St. Louis This article is an editorial note submitted to CCR. It has NOT been peer reviewed. Authors take full responsibility for this article’s technical content. Comments can be posted through CCR Online. ABSTRACT to experiment with production traffic, which have created an This whitepaper proposes OpenFlow: a way for researchers exceedingly high barrier to entry for new ideas. Today, there to run experimental protocols in the networks they use ev- is almost no practical way to experiment with new network ery day. OpenFlow is based on an Ethernet switch, with protocols (e.g., new routing protocols, or alternatives to IP) an internal flow-table, and a standardized interface to add in sufficiently realistic settings (e.g., at scale carrying real and remove flow entries. Our goal is to encourage network- traffic) to gain the confidence needed for their widespread ing vendors to add OpenFlow to their switch products for deployment. The result is that most new ideas from the net- deployment in college campus backbones and wiring closets. working research community go untried and untested; hence We believe that OpenFlow is a pragmatic compromise: on the commonly held belief that the network infrastructure has one hand, it allows researchers to run experiments on hetero- “ossified”. geneous switches in a uniform way at line-rate and with high Having recognized the problem, the networking commu- port-density; while on the other hand, vendors do not need nity is hard at work developing programmable networks, to expose the internal workings of their switches.
    [Show full text]
  • July 18, 2012 Chairman Julius Genachowski Federal Communications Commission 445 12Th Street SW Washington, DC 20554 Re
    July 18, 2012 Chairman Julius Genachowski Federal Communications Commission 445 12th Street SW Washington, DC 20554 Re: Letter, CG Docket No. 09-158, CC Docket No. 98-170, WC Docket No. 04-36 Dear Chairman Genachowski, Open data and an independent, transparent measurement framework must be the cornerstones of any scientifically credible broadband Internet access measurement program. The undersigned members of the academic and research communities therefore respectfully ask the Commission to remain committed to the principles of openness and transparency and to allow the scientific process to serve as the foundation of the broadband measurement program. Measuring network performance is complex. Even among those of us who focus on this topic as our life’s work, there are disagreements. The scientific process happens best in the sunlight and that can only happen when as many eyes as possible are able to look at a shared set of data, work to replicate results, and assess its meaning and impact. This ensures the conclusions from the broadband measurement allow for meaningful, data-driven policy making. Since the inception of the broadband measurement program, those of us who work on Internet research have lauded its precedent-setting commitment to open-data and transparency. Many of us have engaged with this program, advising on network transparency and measurement methodology and using the openly-released raw data as a part of our research. However, we understand that some participants in the program have proposed significant changes that would transform an open measurement process into a closed one. Specifically, that the Federal Communications Commission (FCC) is considering a proposal to replace the Measurement Lab server infrastructure with closed infrastructure, run by the participating Internet service providers (ISPs) whose own speeds are being measured.
    [Show full text]
  • Netfpga—An Open Platform for Teaching How to Build Gigabit-Rate Network Switches and Routers Glen Gibb, Member, IEEE, John W
    364 IEEE TRANSACTIONS ON EDUCATION, VOL. 51, NO. 3, AUGUST 2008 NetFPGA—An Open Platform for Teaching How to Build Gigabit-Rate Network Switches and Routers Glen Gibb, Member, IEEE, John W. Lockwood, Member, IEEE, Jad Naous, Paul Hartke, and Nick McKeown, Fellow, IEEE Abstract—The NetFPGA platform enables students and re- searchers to build high-performance networking systems using field-programmable gate array (FPGA) hardware. A new version of the NetFPGA platform has been developed and is available for use by the academic community. The NetFPGA platform has modular interfaces that enable development of complex hardware designs by integration of simple building blocks. FPGA logic is used to implement the core data processing functions while software running on an attached host computer or embedded cores within the device implement control functions. Reference Fig. 1. Photograph of a NetFPGA installed in a Desktop PC. designs and component libraries have been developed for the CS344 course at Stanford University, Stanford, CA, and taught at a series of tutorials held in the United States, United Kingdom, India, China, Australia, and Europe. The open-source Verilog, C, C commercial vendors of high-speed networking equipment Perl, and Java reference design is available for download from the use application specific integrated circuits (ASICs) and/or project website. field-programmable gate arrays (FPGAs) to accelerate the Index Terms—Field-programmable gate arrays (FPGAs), In- switching, routing, and processing of packet data. To be com- ternet, networks, protocols, routing, switches. petitive, students need to understand how these hardware-accel- erated systems operate. Using the NetFPGA platform, students can build and prototype their own hardware-accelerated net- I.
    [Show full text]
  • Nick Mckeown Academic Employment Current Research Interests
    Last updated: October 10, 2017 Nick McKeown Departments of Computer Science Tel: (650) 725­3641 & Electrical Engineering Gates 344 Email: [email protected] Stanford University Stanford, CA 94305­9030 http://www.stanford.edu/~nickm Academic Employment Stanford University ● Kleiner Perkins, Mayfield, Sequoia Professor of Engineering (2012­ ) ● Professor of Electrical Engineering and Computer Science (2010­ ) ● Faculty Director, Open Networking Research Center (2012­2016) ● Faculty Director, Clean Slate Design for the Internet (2006­2012) ● Associate Professor of Electrical Engineering and Computer Science (2002­2010) ● Assistant Professor of Electrical Engineering and Computer Science (1995­ 2002) Current research interests Software­defined networks (SDN), programmable networks, languages for expressing forwarding behavior, net­neutrality and personalized networks. Academic Background Place of Study Degree Dates University of California, Berkeley PhD May 1995 Electrical Engineering and Computer Science University of California, Berkeley MS May 1992 Electrical Engineering and Computer Science University of Leeds, England BEng May 1986 Electrical and Electronic Engineering Phd Thesis: Scheduling Cells in an Input­Queued Cell Switch. Adviser: Professor Jean Walrand, University of California, Berkeley. Last updated: October 10, 2017 Other Organizations P4 Language Consortium ( P4.org) , Board Member (2014­) Barefoot Networks Inc, Co­Founder, Chairman and Chief Scientist (2013­) Open Networking Lab (ON.Lab), Board Member (2011­) Open Networking Foundation (ONF), Co­Founder and Board Member (2010­) Nicira Networks Inc, Co­Founder and Board Member (2007­2012; Acquired by VMware) Nicira was one of the first “software­defined networking” (SDN) companies and invented the concept of “network virtualization”. Nemo Systems Inc, CEO and Co­Founder (2003­2005; Acquired by Cisco) “Network Memory” saves networking companies hundreds of millions of dollars per year on high price SRAMs for packet buffering and event counters.
    [Show full text]
  • A Network in a Laptop: Rapid Prototyping for Software-Defined
    A Network in a Laptop: Rapid Prototyping for Software-Defined Networks Bob Lantz Brandon Heller Nick McKeown Network Innovations Lab Dept. of Computer Science, Dept. of Electrical Engineering DOCOMO USA Labs Stanford University and Computer Science, Palo Alto, CA, USA Stanford, CA, USA Stanford University [email protected] [email protected] Stanford, CA, USA [email protected] ABSTRACT 1. INTRODUCTION Mininet is a system for rapidly prototyping large networks Inspiration hits late one night and you arrive at a on the constrained resources of a single laptop. The world-changing idea: a new network architecture, ad- lightweight approach of using OS-level virtualization fea- dress scheme, mobility protocol, or a feature to add to tures, including processes and network namespaces, allows a router. With a paper deadline approaching, you have it to scale to hundreds of nodes. Experiences with our ini- a laptop and three months. What prototyping environ- tial implementation suggest that the ability to run, poke, and ment should you use to evaluate your idea? With this debug in real time represents a qualitative change in work- question in mind, we set out to create a prototyping flow. We share supporting case studies culled from over workflow with the following attributes: 100 users, at 18 institutions, who have developed Software- Defined Networks (SDN). Ultimately, we think the great- Flexible: new topologies and new functionality est value of Mininet will be supporting collaborative net- should be defined in software, using familiar lan- work research, by enabling self-contained SDN prototypes guages and operating systems. which anyone with a PC can download, run, evaluate, ex- Deployable: deploying a functionally correct pro- plore, tweak, and build upon.
    [Show full text]
  • A Draft Syllabus [PDF]
    CSci551 Syllabus|FA2020, Monday/Wednesday Section John Heidemann August 5, 2020 Class meets Monday and Wednesday, 10am to 11:50pm, beginning August 24 and ending December 7. There is no class on September 7 (Labor day) nor on November 28 (Thanksgiving recess). We will have two short midterms at 10am September 23 and 10am October 28. The date and time of the final is Monday, December 7, 8am{10am. All students are expected to confirm they can make both the midterm and final exams|we do not offer alternative dates. Please note that the undergraduate term is different this year, starting a week before the graduate term and ending at Thanksgiving. You lucky graduate students get non-COVID timing. As of August 5, USC is not opening in-person classes this fall, so at least initially all classes will be on-line. DEN prefers Cisco WebEx as the platform and we will use that. I have an interactive lecture style and will do my best to adapt it to online class|I strongly encourage students to attend class synchronously and be prepared to comment during class to get the most out of class. Changes: This syllabus may be updated over the semester. The most recent version can always be found at the class Moodle site. 2020-08-05: no changes yet Obtaining class papers: All class papers are available from the CSci551 Moodle site (described below) in PDF format. Because they are copyrighted they are available only for classroom use. The Moodle site is only available to students with class-specific accounts.
    [Show full text]
  • Netfpga – an Open Platform for Teaching How to Build Gigabit-Rate Network Switches and Routers
    NetFPGA – An Open Platform for Teaching How to Build Gigabit-rate Network Switches and Routers Glen Gibb, John W. Lockwood, Jad Naous, Paul Hartke, and Nick McKeown Abstract—The NetFPGA platform enables students and researchers to build high-performance networking systems using Field Programmable Gate Array (FPGA) hardware. A new version of the NetFPGA platform has been developed and is available for use by the academic community. The NetFPGA platform has modular interfaces that enable development of complex hardware designs by integration of simple building blocks. FPGA logic is used to implement the core data processing functions while software running on an attached host computer or embedded cores within the device implement control functions. Reference designs and component libraries have been developed for the CS344 course at Stanford University and an open-source Verilog reference design is available for download from the project website. Index Terms—Field programmable gate arrays, Internet, networks, protocols, routing, switches. I. INTRODUCTION High performance network switches and routers enabled the rapid growth of the Internet. Gigabit Ethernet switches are widely deployed to interconnect computers in Local Area Networks (LANs). Multi-Gigabit/second links are used to transport Internet Protocol (IP) packets across Wide Area Networks (WANs). At most universities, students only learn to build networking systems with software. Students who take a hands-on course in computer networking write software programs that send and receive packets through user-space sockets. Students who take advanced courses write software for the kernel that interfaces directly with a Linux operating system. While systems implemented with software may be able to send and receive some of the packets to and from a Gigabit/second Ethernet line card, software alone is not suitable for switching, routing, and processing all of the traffic that that appears on high-speed networks.
    [Show full text]
  • From Ethane to SDN and Beyond
    From Ethane to SDN and Beyond Martín Casado Nick McKeown Scott Shenker Andreessen Horowitz Stanford University University of California, Berkeley [email protected] [email protected] [email protected] This article is an editorial note submitted to CCR. It has NOT been peer reviewed. The authors take full responsibility for this article’s technical content. Comments can be posted through CCR Online. ABSTRACT and expensive, and internally they were based on old engineering We briefly describe the history behind the Ethane paper andits practices and poorly defined APIs. ultimate evolution into SDN and beyond. Yet, most customers only used a handful of these features. One approach would have been to completely redesign the router hard- CCS CONCEPTS ware and software around cleaner APIs, better abstractions, and modern software practices. But time-to-market pressures meant • Networks → Network architectures; Network types; they couldn’t start over with a simpler design. And while in other industries startups can enter the market with more efficient ap- KEYWORDS proaches, the barrier to entry in the router business had been made Software Defined Networks (SDN) so tall that there was little chance for significant innovation. Instead, the router vendors continued to fight problems of reliability and security brought on by overwhelming complexity. The research 1 THE ETHANE STORY community, sensing the frustration and struggling to get new ideas SDN is often described as a revolutionary architectural approach adopted, labeled the Internet as “ossified” and unable to change. In to building networks; after all, it was named and first discussed in response, they started research programs like NewArch [7], GENI the context of research.
    [Show full text]
  • Pablo Molinero-Fernández Ph.D
    Pablo Molinero-Fernández Ph.D. in Electrical Engineering Batalla de Garellano, 26 +34-91 307 9194 / +34 665 055 824 Madrid 28023 http://klamath.stanford.edu/~molinero/ Spain EDUCATION Stanford University Stanford, California 10/96-6/03 Ph.D. in Electrical Engineering Dissertation: Circuit Switching in the Internet. Advisor: Prof. Nick McKeown. This study discusses the advantages and disadvantages of using circuit switching in the core of the Internet from both a technological and economical point of view. It proposes two network architectures that integrate a circuit-switched backbone with the rest of the Internet in an evolutionary manner. One approach uses fine-grain, lightweight circuits, the other coarse, heavyweight circuits. Reading committee: Professors Nick McKeown, Balaji Prabhakar and Nick Bambos. 9/95-6/96 Master of Science in Electrical Engineering with a specialization in Computer Networking. Advisor: Prof. Fouad Tobagi. École Nationale Supérieure des Télécommunications (ENST ) Paris, France 9/92-7/94 French Advanced Telecommunications Engineer ("Ingénieur des Télécommunications") Escuela Técnica Sup. Ingenieros Telecomunicación (ETSIT) Universidad Politécnica de Madrid, Spain 10/88-7/94 Spanish Advanced Telecommunications Engineer ("Ingeniero Superior de Telecomunicación") These two degrees were obtained in a six-year, double-degree program involving studies at the two graduate-level engineering institutes for telecommunications. Spent the last two years at ENST Paris (92-94), specializing in design and architecture of computer systems, and the first four years at ETSIT Madrid, with a specialization in computer networks and microelectronics. Universidad Nacional de Educación a Distancia Madrid, Spain 8/90-5/95 “Licenciado” in Physics. Five-year program culminating in the equivalent of a Masters of Science.
    [Show full text]
  • George Varghese
    GEORGE VARGHESE Microsoft Research 1288 Pear Avenue 858{335{6996(Cell) Mountain View, CA 94043 650{963{9609(Home) Net: [email protected], www-cse.ucsd.edu/users/varghese/ EDUCATION Massachusetts Institute of Technology, Ph.D. (Computer Science), Feb 1993. North Carolina State University, Raleigh, M.S. (Computer Studies), Aug 1983. Indian Institute of Technology, Bombay, B.Tech (Electrical Engineering), Aug 1981. EXPERIENCE Academic/Research: Aug 2012 - present: Partner and Principal Researcher. Microsoft Research. Network Verification, Geo-distributed analytics Aug 2011 - June 2012: Academic Visitor, Yahoo! Research, Santa Clara Designing a content markeplace, coordination tools. Aug 2010 - July 2011: Distinguished Visitor, Department of Computer Science, Stanford Univer- sity. Network Verification, Abstractions for Genomics. Aug 2000 - Dec 2012: Full Professor of Computer Science, ending at Step 6 Measurement Algorithmics, Security Algorithmics Sept 1993 - Aug 99: Associate Professor/Full Professor of Computer Science, Washington Uni- versity at St. Louis. Network Algorithmics, Self-stabilization Non-academic: May 2005 - Aug 2012: Technical Leader, ISBU for 1 year, then Consultant, Cisco Systems Inc. Helped transition the NetSift technology to a 20 Gbps chip called Hawkeye May 2004 - May 2004: President, CTO, and Co-Founder of NetSift Inc. NetSift was a UCSD spinoff that developed automated techniques for learning and detecting attack signatures. Net- Sift was acquired by Cisco in May 2005 Aug 1983 - Aug 1993: DECNET Architecture and Development, Digital, Littleton, MA. Various positions, ending as a Principal Engineer. Network Architect for DEC's Corporate DEC's next generation network and wrote specification for DEC's Bridge Architecture, later adopted by the IEEE 802.1 committee Technical Advisory Boards: Memoir Memory Systems (acquired by Cisco); Sanera (acquired by McData), Jibe (acquired by Citrix), and SwitchOn (acquired by PMC-Sierra).
    [Show full text]
  • Professor Nick Mckeown Freng
    PROFILE EVOLVING THE INTERNET Professor Nick McKeown FREng He may have given the world the technology that speeded up the internet, but in his next move, Professor Nick McKeown FREng plans to replace those networks he helped create. It would be hard to think of anyone more MIND MADE UP appropriate to interview over an internet Although McKeown did what he was told link between Sussex and California than by his new employers, this had not always Professor Nick McKeown FREng. After all, his been the case. Had he heeded his careers PhD research findings delivered a tenfold master at school, he would have dismissed increase in the speed of routers, which engineering altogether. When McKeown let enabled the internet to handle the traffic on that he was considering an engineering created by Skype and other services. education, inspired in part by his father, Back in the 1980s, McKeown was assigned Professor Pat McKeown FREng, an engineer his first research tasks at the Hewlett Packard and entrepreneur, he recalls that the careers (HP) Labs near Bristol. At the time, he hoped master’s response was: “You are too smart to work on artificial intelligence or computer Nick McKeown gives a TED talk at a conference in to be an engineer. You should go into Monterey in 2006 © Robert Leslie architecture, the popular topics of the day. something creative.” McKeown laughs and However, HP decided to focus McKeown’s says: “From that moment my mind was time and research resources on internet made up. I decided: ‘Right, I am going to be router architecture, which quickly became an engineer!’ one of the hottest areas of technology.
    [Show full text]
  • The Nanopu: a Nanosecond Network Stack for Datacenters
    The nanoPU: A Nanosecond Network Stack for Datacenters Stephen Ibanez, Alex Mallery, Serhat Arslan, and Theo Jepsen, Stanford University; Muhammad Shahbaz, Purdue University; Changhoon Kim and Nick McKeown, Stanford University https://www.usenix.org/conference/osdi21/presentation/ibanez This paper is included in the Proceedings of the 15th USENIX Symposium on Operating Systems Design and Implementation. July 14–16, 2021 978-1-939133-22-9 Open access to the Proceedings of the 15th USENIX Symposium on Operating Systems Design and Implementation is sponsored by USENIX. The nanoPU: A Nanosecond Network Stack for Datacenters Stephen Ibanez, Alex Mallery, Serhat Arslan, Theo Jepsen, Muhammad Shahbaz?, Changhoon Kim, and Nick McKeown Stanford University ?Purdue University Abstract from when a client issues an RPC request until it receives a We present the nanoPU, a new NIC-CPU co-design to response) for applications invoking many sequential RPCs; accelerate an increasingly pervasive class of datacenter appli- (2) the tail response time (i.e., the longest or 99th %ile RPC cations: those that utilize many small Remote Procedure Calls response time) for applications with large fanouts (e.g., map- (RPCs) with very short (µs-scale) processing times. The novel reduce jobs), because they must wait for all RPCs to complete aspect of the nanoPU is the design of a fast path between the before continuing [17]; and (3) the communication overhead network and applications—bypassing the cache and memory (i.e., the communication-to-computation ratio). When com- hierarchy, and placing arriving messages directly into the CPU munication overhead is high, it may not be worth farming out register file.
    [Show full text]