During recent years, the Internet has witnessed rapid growth of streaming applications(audio and video) despite the lack of support for quality of service(QoS). Many of these applications do not perform congestion control effectively. Thus there is significant concern about the effects on co-existing well-behaved traffic and the potential for congestion collapse. However, performing congestion control could result in wide and random variations in transmission rate whereas streaming applications require sustained average transmission rate. To cope with these conflicting requirements, we propose that streaming applications perform "quality adaptation" on-the-fly as available bandwidth changes during a session.

In this talk, we will present an end-to-end architecture suited for unicast playback of layered-encoded stored multimedia streams over the Internet. We focus on congestion control and quality adaptation, and describe how they are reconciled in our architecture. The architecture exhibits a TCP-friendly behavior by adopting the Rate Adaptation Protocol (RAP) for end-to-end congestion control. Additionally, it employs a layered framework for quality adaptation to maximize perceptual quality while minimizing rapid, disturbing changes in quality of the delivered stream as the available bandwidth changes. Furthermore, the quality adaptation mechanism provides a tuning parameter that allows the server to trade short-term improvement for long-term smoothing of delivered quality.

The average delivered quality of the end-to-end approach is limited to the available bandwidth between two end points. Proxy caches perfectly complement our end-to-end architecture to overcome this limitation and maximize the delivered quality of popular streams. We describe a fine-grain replacement algorithm for a proxy caching mechanism of layered-encoded multimedia streams as well as a pre-fetching scheme to smooth out the variations in quality of a cached stream. The interaction between the pre-fetching and the replacement algorithm results in the state of the cache converging to the optimal state such that the quality of a cached stream is proportional to its popularity, and the variations in quality of the cached stream are inversely proportional to its popularity.

He has been a member of Netstation, Virtual Internet Testbed(VINT) and Scalable Personal Tele-conferencing(SPT) projects at the USC Information Sciences Institute(ISI) during last three years.