| The emergence of the Internet as a pervasive communication medium has fueled a dramatic convergence of voice, video and data on this digital information infrastructure. Demand for streaming media is growing exponentially, and the Internet's infrastructure often fails to meet this demand. Video streams typically have high bandwidth requirements even when compressed (e.g., 4–6 Mbps average for MPEG-2), exhibit burstiness on multiple time scales, and can be relatively long lived, making it expensive to deliver multimedia content. In addition, the best-effort service model of traditional IP networks undermines the effort to stream high-quality video across the Internet in real time.; In this thesis, we explore a number of key challenges associated with delivering streaming multimedia media from a server to client(s) across a heterogeneous network such as the Internet, and develop effective solutions to these problems. In particular, we develop and evaluate services that, running on origin servers and on proxy servers located at the edge of, and within, the network, can facilitate high quality multimedia streaming to clients, while maintaining low end-to-end resource requirements, and making efficient use of these resources.; The main contributions of this thesis are: (1) An online bandwidth smoothing technique for reducing the bursty transmission band-width requirements of streaming variable-bit-rate (VBR) video. (2) A network proxy-based prefix caching scheme that caches initial portions of popular videos at edge proxies close to clients. The technique enables clients to experience high-quality streaming video playback with low playback startup latencies. (3) An application-level multicast scheme, integrating bandwidth smoothing and temporal caching, for reducing the network transmission bandwidth requirements when streaming a VBR video to multiple heterogeneous clients across a heterogeneous network like the Internet. (4) Patching and periodic broadcast schemes for reducing the transmission bandwidth associated with streaming a popular video to multiple asynchronous clients, by allowing clients arriving at different times to share all or part of a single transmission. |
