Router-based Congestion Control -- Our previous congestion control work focused on end-to-end feedback and adaptation. Here we are implementing an end-to-end connection as a series of end-system-to-router or router-to-router connections and are recursively deploying end-system adaptations between each pair of endpoints of the logical connections. The goal is to demonstrate a framework for utilizing quality-of-service-based transmission where it exists in the network and to ameliorate the effects of congestion where such services do not exist. This work is the dissertation research of Mark Parris, who held the Intel Graduate Student Fellowship during 1997-98.

TCP Congestion Control Dynamics and Continuous Media -- Given that real-time multimedia traffic eschews TCP, a concern within the Internet community is that best-effort congestion control schemes for multimedia traffic achieve their performance by "stealing" bandwidth from TCP connections that will stop transmitting upon sensing congestion. Here we are investigating the interplay between modern TCP congestion control mechanisms and those used for continous media streams.

To assist with these two research efforts, Intel-based computers donated through the Intel Technology for Education 2000 program have been used to upgrade the facilities in our Multimedia Networking Lab and in some offices. The new machines have enabled us to build a networking infrastructure for our research. We have constructed a network and attached machines to it in order to generate network traffic. This, in turn, creates a testbed for our work on real-time transmission of data over a network. In addition, the lab machines support courses related to this research, such as the new course on Advanced Networking: Internet Architecture and Performance which is being offered for the first time.

Network Support for Distributed, Interactive, Virtual Laboratories -- This project is a collaboration with the nanoManipulator group to distribute their system across campus and wide-area networks. By considering the distribution of a virtual reality system, this project attempts to understand how the lessons learned for the management of audio and video data on the Internet scale to address the requirements for other continous media types, such as display lists, user's position and orientation tracking reports, and instrument control message streams. Our research considers the trade-offs among computer-human interfaces, network infrastructure for real-time communications, and costs of replicating special-purpose hardware. The goal is to understand which system structures and user interaction models are required for achieving useful, cost-effective results in a distributed laboratory. We also seek to enable and foster scientific collaboration and to better utilize expensive, special-purpose instruments. Our first experiment was the distributed nanoManipulator project, in which we took the nanoManipulator to Orange High School in Hillsborough, N.C., and transmitted data back and forth across the Internet from the scanned probe microscope in the UNC Physics Department to students at the high school who were running tests on the nanoManipulator. We are now working on network support for systems distributed to other research labs on the UNC campus (Chemistry, Gene Therapy, and Physics), to the National Institute for Environmental Health Sciences in Research Triangle Park, N.C., and to Professor Don Brenner's Materials Sciences lab at NC State University.

For further description of multimedia networking research, see the Research in Multimedia Systems research page and the Distributed and Real-Time Systems research page.