Large Scale Communication Networks

Large-Scale Communication Networks: Topology, Routing, Traffic, and Control

March 18 - 22, 2002

Schedule and Presentations

Pictures

Organizing Committee:

David Donoho (Stanford University)
John Doyle (California Institute of Technology)
Frank Kelly (Stanford University)
Walter Willinger (AT&T)

Scientific Content

This workshop will focus on the large-scale nature of the Internet and how to account for it adequately when studying core Internet issues such as topology, routing, traffic, performance, and control. Standard theory poorly addresses many of these issues. The feedback in routing and congestion control challenges the standard Shannon abstractions for viewing sources and channels, and the decentralized nature of the control challenges standard control theory. What is needed is a more unified information and control theory for Internet-like networks. One approach to develop such a theory involves the use of a combination of large-scale network measurement infrastructures and large-scale network simulation engines. A central problem in faithfully simulating large-scale networks is that packet level descriptions scale poorly; while there exist more abstract models using, say, flows, the errors made in such approximations are not understood.

The central mathematical questions to be addressed are: 
- a more unified treatment of communication and control
- multiscale, multi-resolution representation/analysis/visualization/models
of large-scale networks
- graph-theoretic treatment of different aspects of network topology
- dynamical systems models for large-scale TCP/BGP dynamics.

The first two become even more acute when moving out of the Internet context (e.g., see other two workshops). There are also many statistics-related issues that are relevant for these topics; e.g., how to infer/determine/characterize spatio-temporal traffic patterns and topology dynamics from network-wide measurements; how to combine traffic and topology characterization and model the spatio-temporal dynamic for routing policy and congestion control.

A poorly posed but challenging and -- in view of the other workshops, where we want to integrate and extend networking to converge with a number of other types of networks -- interesting question is “what makes a good protocol stack?” We take TCP/IP so much for granted that it has become a law of nature. At the same time, routing protocols such as BGP or OSPF have attracted considerable less attention, and the need for gaining a sound understanding of routing-related issues is being realized only now. What would a theory of protocols look like?

Currently, there are no general principles of protocol design, and it is not generally considered a serious research problem, yet it could be crucial. This workshop seeks to address these issues. Within the framework of workshop 1, there will be three miniworkshops.

A: Mini-Workshop on "Internet Congestion Control"

Date: Monday/Tuesday morning, March 18-19, 2002

Organizers:

Frank Kelly (Cambridge/Stanford, chair)
Steven Low (Caltech)
R. Srikant (UIUC)

This miniworkshop will be devoted to congestion control in large-scale networks and the nature of Internet congestion. Explicit Congestion Notification, becoming available in the Internet, has provided a substantial impetus to work on congestion control, and exciting progress has been made recently towards a theoretical understanding of TCP, both of its equilibrium, using optimization theory, and its dynamics, using control theory. In parallel there have been important progress on mathematical models that address the statistical relationship between traffic, capacity, routing and realized performance.

Confirmed speakers:

Eyad Abed (Univ. Maryland)
Francois Baccelli (ENS/INRIA)
Hari Balakrishnan (MIT)
Christophe Diot (Sprint Labs)
Ramesh Johari (MIT)
Peter Key (Microsoft)
Srisankar Kunniyur (Univ. Pennsylvania)
Steven Low (Caltech)
Laurent Massoulie (Microsoft)
Vishal Misra (Columbia)
Teunis Ott (NJIT)
Fernando Paganini (UCLA)
Jim Roberts (France Telecom)
Sanjay Shakkottai (UIUC)
Glenn Vinnicombe (Cambridge/Caltech)
Milan Vojnovic (EPFL)

B: Mini-Workshop on "Internet-wide Measurements"

Date: Tuesday afternoon/Wednesday/Thursday morning, March 19-21, 2002

Organizers:

Paul Barford (Univ. Wisconsin, chair)
Albert Greenberg (AT&T Labs-Research)
Stefan Savage (UCSD)

This mini-workshop brings together speakers from a broad spectrum of networking disciplines to discuss issues and directions in Internet measurements. The workshop will focus in two general areas:

1) measurement tools and infrastructures, and 2) measurements in operational networks which provide a foundation for management and troubleshooting. The workshop program has been developed to present the state of the art in network measurement, and to explore limitations and difficulties of using Internet measurements in support of research or operational functions. Questions which should be answered in this workshop include:

-- What is the scope of current Internet measurement activities?

-- What are the challenges in collecting and analyzing Internet measurement data?

-- What data and techniques are needed to manage and troubleshoot wide area networks?

This workshop will naturally cover aspects of the topics in the first and third mini-workshops, however it will focus more specifically on the tasks of collecting, interpreting, and using Internet measurement data.

Speakers

Contact Us:

Institute for Pure and Applied Mathematics (IPAM)
Attn: CNTOP
460 Portola Plaza
Los Angeles CA 90095-7121
Phone: 310 825-4755
Fax: 310 825-4756
Email: ipam@ucla.edu
Website: http://www.ipam.ucla.edu/programs/cntop/