Full Abstract

This tutorial aims to cover the uses of NetFlow to monitor an SP' infrastructure, from the export, router-wise, to the collect and analysis, from a security perspective. While there is certainly a great deal of papers on the subject, there is room - and need, or so I hope - for a comprehensive coverage of NetFlow from theory to practical operations. The questions regularly popping-up in the operators' forums about NetFlow uses, as well as the lack of visibility most operators have regarding attacks against their infrastructure may be good testimonials to support such a tutorial.

Speakers
Yann Berthier

Full Abstract

I'm trying to put together a BOF to be held during the NAOG 39 meeting (Feb 4-7), that looks at when current router/switch architectures run out of TCAM or SRAM on the line-cards (depending on flavor) to hold the fib, and potentially what's being done about it. Part of the whole global routing scalability problem that's been fodder for a couple NANOG meeting now hinges on having the data you need in the fib. Comparatively buying more ram for or scaling your rp is a problem, but not immediately an intractable one by comparison. I'm aiming to do this as a BOF rather than as part of the main program in the hope that vendors will be more free to talk specifically about their own products in a way that is not always appreciated in the main meeting, Feedback has been positive on this approach from the Program Committee... I would expect that this topic would have a significant audience with the NANOG crowd. If I've reached the wrong person in your organization can you please help me find someone who might be interested because I think this topic has serious implications for operators and will probably shape purchasing decisions among other things over the coming year. Currently I've solicited a couple router/switch vendors and a few (two) have already gotten back to me. I intend to also solicit presentations from the operator community, because I know for a fact that people are doing interesting things when confronted with this problem. A large regional ISP that I know in a non-principally-english speaking country is continuing to buy routers with 200k fib entries for example and counting on aggressive filtering and a potentially dramatically incomplete view of the internet to be sufficient.

Speakers
Joel Jaeggli, Nokia

Full Abstract

This tutorial gives an overview of how IP multicast and (MPLS) multipoint technologies can be used in IPTV solutions and beyond. The presentation first sumamrizes such solutions and the resulting strategies to consider when deploying multipoint technologies. It then continues to present an architectural overview highlighting key functions of multipoint IPTV solution. In the remainder of the session those key functions are reviewed SSM, native IP multicast vs. mLDP and RSVP-TE/P2MP tree building, resiliency for multipoint (convergence, fast reroute, dualstream), Path selection for cost optimization, admission control and channel change issues and solutions.

Speakers
Toerless Eckert, Cisco Systems

Full Abstract

Arbor Networks

Full Abstract

Speakers
Steve Feldman, CNET
Sylvie LaPerrière, Teleglobe - VSNLI
Jon Nistor, TorIX

Full Abstract

A succession of major earthquakes in the Luzon Strait on December 26, 2006 caused 7 out of 8 undersea ring-protected subsea cable systems to fail, interrupting the majority of voice, data and Internet communications in Southeast Asia. The physical subsea routes required to build an Internet backbone from the USA to Southeast Asia are depicted. Natural catastrophes remind service providers that physical route diversity planning over multiple subsea cable assets laid in close proximity (Luzon Strait) might not be enough. The severity of damages is shown and the adversity of cable repair conditions and timeline explained. The current growth of Internet traffic within the Asia Pacific region is creating economic incentives to build future cables on a westward route to Europe thus reducing dependency on the Luzon strait route.

Speakers
Sylvie LaPerrière, Teleglobe - VSNLI
Sylvie LaPerrière is Director of Peering and Commercial Operations at Teleglobe, a VSNL International company. She is responsible for overseeing the company’s Internet expansion and market management worldwide. She is also Chair of the Teleglobe Peering Committee. Ms. LaPerrière joined the company in 1993 and launched Teleglobe’s first Internet service in 1995. Since 2002 she has been leading Teleglobe’s global Internet expansion into new markets. Ms. LaPerrière has 13 years of extensive experience in product management for telecommunication services. Ms. LaPerrière earned her Bachelor of Commerce in Marketing and Information Systems from École des Hautes Études Commerciales de Montréal.

Full Abstract

Six large quakes hit Taiwan in rapid succession on Dec 26, 2006. As a result, at least six major cables were severed causing major disruption to Internet routing. Using internet routing data from 150+ sessions, we look at the effects of these events on routing. We look at instability and outages both, and find that over 19K networks were impacted by the event, including, at peak level, almost 4k networks unreachable. We find that the event is more complicated, and more interesting, than some previous disasters because of the compound nature of the underlying failure (six cables). In particular, the onset of the event is less clearly tied to the timing of the quakes, and the recovery is more complicated as well. Finally, we look at the winners and the losers in the event, from an autonomous system perspective and from a country perspective.

Speakers
Alin Popescu, Renesys Corporation
Alin Popescu is a member of the engineering team at Renesys. His specialties include implementating statistical and learning algorithms and developping system architectures for BGP data analysis. Before joining Renesys, Alin earned a Ph.D. in Computer Science from Dartmouth College.

Todd Underwood, Renesys Corporation
Todd Underwood is in charge of operations, and peering for Renesys. Before that he was CTO of Oso Grande, a New Mexico ISP. He has a background in systems engineering and security and has worked on a variety of systems architecture and scalability problems. Todd has presented work related to Internet routing dynamics and relationships at NANOG and various peering forums (LINX, S&D, NOTA). Todd has a BA in Philosophy from Columbia University and a MS in Computer Science from University of New Mexico.

Earl Zmijewski, Renesys Corporation
Earl Zmijewski is responsible for Renesys's professional services. He has nearly 20 years of experience encompassing scientific computing and most areas of IT, with particular emphasis on networking and security. Before Renesys, Earl was IT Director at Fluent Inc., a computational fluid dynamics software company, where he was instrumental in establishing new offices throughout the US, Europe and Asia and in the promotion and implementation of Linux clustering technologies. He was also principal architect in the design of Fluent's networks and Internet security posture. Before that, Earl held various academic positions at Cornell University, University of California, and James Madison University. Earl has a PhD and MS in Computer Science from Cornell University and an MS and BA in Mathematical Sciences from The Johns Hopkins University.

Full Abstract

Since the controversy over "stateless anycast" refuses to go away--even if it's kept alive mostly by one person--we thought it would be interesting to reexamine some of the data we presented two years at NANOG in our "Life and Times of J Root" paper. In particular, we will again look at all anycast instances of J root to see which source IP addresses show up at multiple locations and how often. We will perform additional analysis to attempt to explain the results, whatever they may be. We will also reexamine the question we considered two years ago, "Who is querying j.root-servers.net's old IP address and why?", to see if the nature of the queriers has changed.

Speakers
Piet Barber
Mark Kosters
Matt Larson

Full Abstract

One of the current impediments in advancing the state of the art in Internet security research is the lack of scalable network tracing platform available to researchers. Given the complexity of today's attacks, packet-level tracing tools are inadequate; instead, today's tracing platforms must reconstruct traffic into application state and inspect it for suspicious or deviant behavior. Unfortunately, no such open-source network tracing software is available to researchers. In our project, we are building a highly scalable, open-source network tracing platform that offers adequate privacy and anonymity guarantees to the users whose traffic is monitored. Our tracing infrastructure is reconstructing traffic across several layers, from network to the application layer. We reassemble IP fragments into IP packets, TCP segments into TCP conections, and TCP connections into HTTP transactions. To protect users' anonymity, our platform does not store any unanonymized data in stable storage. Instead, all work (including capturing and reassembly) is done in volatile memory at line speeds. In our presentation at NANOG, we focus on the privacy requirements for tracing the network traffic of a large Internet user population. We start by describing a list of attacks possible when collecting application-level information by network tracing. We argue that the only way to mitigate the privacy implications of all these attacks is to never store any un-anonymized data to stable storage when tracing. We then present a high-level overview of our open-source network tracing infrastructure. Our goal is to get the NANOG's community feedback on our anonymization and privacy protocol as well as their input in the design of our monitoring platform.

Speakers
Troy Ronda, Computer Science, University of Toronto
Stefan Saroiu, Computer Science, University of Toronto
Stefan Saroiu joined the Computer Science at the University of Toronto in 2005 after a brief hiatus at Amazon.com. Stefan received his Ph.D. in 2004 from the University of Washington where he worked with Steve Gribble and Hank Levy. Stefan's research interests span the range from operating systems to networking and distributed systems.

Full Abstract

A recent statement by AT\&T CEO Ed Whitacre sparked considerable fear in the public that the Internet may not be open any more: the ISPs dictate which sites/applications flourish and which flounder. The statement triggered the heated debate on net neutrality and ignited the battle to enact net neutrality legislation. However, by the date of writing, all attempts to pass net neutrality laws have failed. This paper states our proposition on net neutrality: ISPs should not be able to discriminate against packets based on contents, application types, or packet sources or destinations that are not their own customers; but they are eligible to offer differentiated services to their customers. We present a technical design that aims to achieve this definition of net neutrality. Our design prevents an ISP from deterministically harming an application, a competing service, or singling out an individual innovator for extortion.

Speakers
Xiaowei Yang, UC Irvine
Xiaowei Yang is an assistant professor in the Department of Computer Science at the University of California, Irvine. She received a Ph.D. in Computer Science from MIT in 2004, and a B.S. in Electronic Engineering from Tsinghua University, Beijing, China in 1996. Her research area is in networks and distributed systems, with an emphasis on protocol and architecture design.

Full Abstract

Many studies show that when Internet links go up or down, the dynamics of BGP may cause several minutes of packet loss. The loss occurs even when multiple paths between the sender and receiver domains exist, and is unwarranted given the high connectivity of the Internet. Instead, we would like to ensure that Internet domains stay connected as long as the underlying network is connected. We present R-BGP, which ensures continuous path availability with minimal overhead by pre-computing a few strategically chosen failover paths. R-BGP provably guarantees that a domain will not experience any disconnectivity if a policy-compliant path exists after convergence. Surprisingly, this can be done using a few simple and practical modifications to BGP, and, just like BGP, requires announcing only one path per neighbor. Simulations using the current Internet graph show that R-BGP reduces the number of domains transiently disconnected by a down link, from 15% to zero.

Speakers
Nate Kushman, MIT
Nate Kushman is pursing a PhD at MIT. His current research is in the area of reliable inter-domain routing. He recieved his B.S. and MEng degrees in Computer Science, in 1998 from MIT. He was a co-founder and Chief Technology Officer at UBUBU for 2 years, and then a senior architect at Akamai for four years until returning to MIT in 2005.

Full Abstract

Speakers
Chris Street

Full Abstract

Speakers
Philip Smith

Full Abstract

Speakers
John Kristoff

Full Abstract

Peering BOF Agenda Once again we will gather in the round to discuss and debate issues of relevance to the peering community. Peering vs. transit, peering contact information, capacity planning, transit purchase terms, international peering, and a historical view are all on the agenda during this double-slot NANOG Peering BOF session. Transit Survey - Dan Golding - Tier 1 Research - 10 minutes As an alternative to peering, it is helpful for the community to get a rough idea of the current market price for transit. As prices of transit drop, does it still make sense to peer away some of that traffic at the cost of transport and peering infrastructure? Dan has formatted a survey and has agreed to share the results with the group at the next Peering BOF. PeeringDB.com - TBD - 5 minutes The community resource peeringdb.com is a repository for peering contact information to aid in establishing and maintaining peering relationships. Of course, a database of this kind is only as valuable as its information is valid, and contact information becomes stale very quickly. So rather than having each IX host this contact information on their web site, and track or allow it to get out of date, peeringdb.com provides a single location for hosting this out of date information. This talk will provide an overview and walk through of the utility of the information. 100G ethernet status update -- Greg Hankins - Force 10 Networks - 10 minutes As the next wave of Internet Video traffic enters the US Peering Ecosystem, it is critical to plan for the scaling of the underlying infrastructure. Current backbones consist of bundles of 10G transport links which are quickly growing to unmanageable levels. The AMS-IX is saying that 2009 is far too late for 100G to help with the growth of its switching fabric. The IEEE is exploring specifications for 100G ethernet, so this talk will describe the progress to date and project the date of delivery. Open Peering is Dumb Discussion/Debate - 30 minutes This controversial title highlights a couple schools of thought in this community. The term "Open" peer refers to a company that will peer unconditionally with anyone who asks. The term "Selective" peer refers to a company that will peer but only with some prerequisites satisfied (multi-location peering, traffic volume minimum, out:in ratio maximum, etc.) This discussion/debate will explore the divide between these two philosophies. Transit Purchase Parameter Rationality Discussion. Denver Maddux - LimeLight Networks - 20 minutes Transit terms are rife with variables including commits, caps, reverse tiered burst pricing, etc. that some content heavy customers find onerous and problematic. The plea from one customer is to remove these terms and conditions from transit, and Denver will highlight what several of these terms signal to the marketplace. Of course, there are two sides to this discussion/debate, as transit providers need to manage their own backbone bandwidth, commits, ratios, 95th percentile tricks, etc. International Peering Experiences - Sylvie LaPerrière - VSNL - 20 minutes Even through the integration of TeleGlobe into VSNL over the past year, there has been great progress made expanding peering infrastructure into and throughout Asia. Sylvie will share with the group some of the challenges, surprises, trade offs, and observations from this recent experience. A view of Peering from Past to Present - facilitated by Chris Quesada (Switch & Data) -- 45 minutes The purpose of this mini-session is to bring together a panel of past and present peering coordinators to share their experiences in peering. Speakers will share their experiences on the following topics; Changes in Peering Geography, Challenges peering presents, and the Politics of Peering. The topics will be address from both sides to show both past and present perspectives and perhaps lend creditability to old adage of "history repeating it self". Anybody who has ever wondered what it was like to peer in the "Good Ole Days" or just wants to hear how they do it now will be most welcome to attend. -- Panelists: TBD Peering Personals - ALL In the remaining time, we will open the floor to the peering coordinators to introduce themselves, their AS#, their peering policy, peering locations, what they look for in a peer, and why companies should want to peer with them. Details such as transit volumes, traffic ratios, email address, etc. may also be shared. As we break for the evening, peering coordinators use this opportunity to meet interesting peers and hopefully establish additional peering sessions.

Speakers
Bill Norton, Equinix
Bill Norton is Co-Founder and Chief Technical Liaison for Equinix. He focuses on research on large-scale interconnection and ISP peering, and in particular, scaling Internet operations using optical networking. Over the last eight years, Bill has published over a dozen industry white papers and presented his research in a variety of international forums. From 1987 to 1998, he served in several staff and managerial roles at Merit Network, directing national and international network research and operations activities and serving as NANOG coordinator. Bill received a B.A. in Computer Science and an M.B.A. from the Business School at the University of Michigan.

Christopher Quesada, Switch & Data.

Full Abstract

This tutorial covers common problems ISPs have when deploying BGP within their network. It looks at problems with peer establishment, missing routes, inconsistent route selection, and convergence issues. It also looks at real world examples of common errors which are made when deploying BGP, both as iBGP and eBGP, in service provider networks.

Speakers
Philip Smith, Cisco Systems
Philip Smith has been with Cisco Systems since 1998 and is based in Brisbane, Australia. He is a Consulting Engineer, part of the Service Provider Architectures Group in Corporate Development. His role includes working with many ISPs in the Asia Pacific region, specifically in network strategies, technology, design and operations, configuration and scaling. As part of an ISP and Internet education initiative, Philip runs several Routing and Internet Technology Workshops in the Asia Pacific region. He also assists as co-instructor at similar events in many other parts of the world. Philip also is closely involved in regional activities, being chair of the APRICOT Management Committee, chair of APOPS, member of the organising and programme committees for SANOG and PacNOG, as well as chair of APNIC's Routing and Internet Exchange Point Special Interest Groups. Prior to joining Cisco, he spent five years at PIPEX (now integrated into MCI's global network business), the UK's first commercial Internet Service Provider. He was one of the first engineers working in the commercial Internet in the UK, and played a key role in building the modern Internet in Europe.

Full Abstract

This BOF is intended to be for those who either run an IPv6 network or are interested in doing so. It is NOT intended to be a place to argue the pros or cons of IPv6. Various issues to do with operating an IPv6 network will be discussed, including; transition methods, peering, tunnelling, filtering, support, network capacity and future issues.

Speakers
Stewart Bamford, Level 3 Communications
After receiving his B.Sc. in Computer Science from the University of East Anglia in 1995, Stewart started work in the ISP business, launching a small ISP in Norwich, England. In 1996, he joined PIPEX in Cambridge (UK), which was later known as UUNET, WorldCom, and then MCI. In 2004 he moved to Level 3 Communications, where he is currently employed as a Senior Engineer based in London. Stewart was the lead engineer in the deployment of IPv6 on the Level3 (AS3356) network. He is also also a Chartered IT Professional, a Chartered Engineer and a member of the British Computer Society.