STI: Media Aware Congestion Control

University Of Southern California, Los Angeles CA

Investigators

Abstract

The use of media rich networked applications has the potential to greatly enhance learning, collaborations, and scientific exchange and the beginning of widespread deployment of tools for networked collaboration, such as video conferencing, telepresence, distance learning, and voice-over-IP; and the potential of the Internet as an enabler for ad-hoc real-time communication is starting to be realized. This poses a potential threat to the stability of the Internet. This is because most media rich applications use transport protocols that do not respond to network congestion. For example, many systems use UDP/IP transport, due to their requirements for timely delivery, and lack the built in congestion control mechanisms of the traditional TCP/IP protocol. If non-congestion controlled traffic continues to proliferate, there is the real potential for congestion collapse of the network. Media applications ignore the need for congestion control because of the lack of a readily available and standardized way of deploying congestion control for media applications. Congestion control schemes designed for media, such as the TCP-Friendly Rate Control protocol (TFRC) are not integrated into the transport protocols-for example, the Real time Transport Protocol (RTP) and UDP-and hence are difficult to use and deploy. This project addresses the lack of a standardized congestion controlled transport for networked multimedia applications. The approach is to incorporate recent developments in TFRC into the RTP framework, thereby providing a unified standard for adaptive and congestion controlled real-time media transport. This will be backed by an open-source implementation, which we will develop to seed deployment of the standard. (1) TFRC has been designed for streaming media applications- so investigate to what extent is it possible to adapt TFRC to the need of interactive video; (2) study the effects of non-ideal network behavior on media applications and congestion control, and propose enhancements to TFRC that make it robust to network and application deficiencies; and (3) design and develop a solution for congestion controlled transport of real-time media, in the form of an RTP Profile for TFRC congestion control. The widespread dissemination of results through activities in the Internet standards community, and through participation in activities within the Access Grid and Conference XP communities will have deep impact. There is considerable pressure to convert networked multimedia applications to congestion controlled transport protocols. The developing of such a standard transport, should remove a significant deployment hurdle for a number of user groups.