On-demand media streaming has recently gained intensive consideration due to its promising usage in a rich set of Internet-based services such as video on demand, distance learning, media distribution, etc. However, there are still many challenges towards building efficient, scalable, on-demand streaming systems, in this thesis, we propose a novel cost-effective on-demand media streaming architecture for heterogeneous peer-to-peer networks, named BitVampire. The key idea of BitVampire is to

more »

... egate peers' storage and bandwidths to facilitate on-demand media streaming. To achieve this goal, we split published videos into segments and distribute them to different peers. When watching a video, a peer searches the corresponding segments, and then aggregates bandwidths from multiple supplying peers to stream the video. To deploy this architecture in a dynamic heterogeneous peer-to-peer network, three key techniques are used: (1) Given that peers offer different resources and may leave at any time, a media segments distributing algorithm and a caching scheme are proposed, which achieve fast system streaming capacity amplification. (2) An application-level overlay, called Category Overlay, is chosen as the underlying search infrastructure to efficiently find the desired segments. (3) A scheduling algorithm is proposed to aggregate bandwidths from multiple supplying peers and coordinate them to serve one streaming request. We demonstrate the effectiveness of this proposed architecture through extensive simulation experiments on large, Internet-like topologies.