LASR NSF Networking Grant

Multicast is a network service for delivering the same data to multiple destinations. A wide range of emerging network applications - such as multi-participant video conferencing, multi-party interactive games, distance learning, group collaboration, and resource location - can greatly benefit from this service.

We propose to develop congestion control protocols for IP multicast to support scalable and efficient dissemination of layered data to a large number of receivers in heterogeneous, dynamic networks. We propose to investigate the use of an integrated set of three distinct mechanisms - per-group feedback-based transmission adjustment, selective participation, and menu adaptation - to solve the general multicast congestion control problem. These mechanisms will operate at different time-scales and distribute the responsibility of adaptation to different entities in the network. We will experimentally evaluate the scalability and performance of our protocols in a large-scale network environment where the multicast session shares network with other unicast and multicast sessions, receivers join and leave the network, and the bottleneck links and bandwidths fluctuate over time. We propose to implement our multicast congestion control protocol in the network testbed of programmable routers, based on the Internet eXchange Architecture (IXA). The outcome of our research will be a family of algorithms, protocols, and prototype implementations that will significantly advance the state of art in designing multicast congestion control protocols.

The TRAM Project:

Our research under this award is performed in the context of LASR's TRAffic Management in Next-Generation Internet (TRAM) project. Today's Internet provides best-effort service to all applications; the network provides little or no guarantees about the end-to-end performance that applications and users expect. Although Internet deployment of several business-critical applications has revealed the inadequacy of best-effort service, this model has led to the development and deployment of a network core that can scale to vast number of users and hosts. Network designers must now meet the challenge of providing rich network services while also maintaining simplicity of network design and implementation. TRAM explores the tradeoffs between richer service semantics and network implementation complexity.

Key Results and Publications:

Please see the both the TRAM and the new Cocoon project pages for up-to-date progress notes, results, and publications.

Collaborative, Educational, and Outreach Activities:

This NSF award has attracted additional funding, equipment donations, and collaborative activities that have had a profound impact on networking research and education in the University of Texas Department of Computer Sciences.

In the 2000-2001 academic year, Intel provided significant equipment support with the donation of 14 high-end Pentium-III Xeon workstations and an IXIA traffic generator. Further, IBM provided us with a cluster of 12 high-end IBM Netfinity servers. The router development platforms based on Intel's IXP1200 network processors are a unique and defining resource for LASR. Our laboratory is one of only about five facilities in the country with access to this emerging technology. This infrastructure has enriched significantly the courses and research in computer networking at UT.

In November 2001, LASR inaugurated its Intel Internet Exchange Architecture (IXA) lab. We will implement our multicast congestion control protocols on this network testbed of programmable routers.

Related research funding:

Invited lecture:
H. Vin, Network Support for Efficient and Scalable Layered Multicast, Intel Corporation, Portland, Oregon, June 2001.