The Research And Performance Optimization Of Large Scale P2P Streaming Networks Based On Network Coding

Peer-to-peer streaming is a cost-effective approach to deliver live videostreams to a large number of clients, and has become an efficient approach tobroadcast video on the Internet and a lot of P2P streaming applications suchas Coolstreaming[12], PPLive[35], PPStream[36], UUSee[7][34], just to mention afew, are becoming more popular and widely used. The essential advantage oflive peer-to-peer streaming is the mitigation of the load on streaming serversand the scalability of a streaming session even with millions of peers. Byleveraging the client-side resources, a P2P video session can scale to millionsof users with only a very small portion of server bandwidth.Network coding technique can improve the efficiency of P2P livestreaming systems, while it may also bring some extra delays andcomputational overhead. In this paper, we present DirectedPush, a highperformance P2P live streaming system based on network coding. InDirectedPush, the original video is encoded into a number of codedsub-streams with improved content diversity. Peers are organized into atwo-layer overlay and use an efficient push algorithm to deliver the codedsub-streams. The proposed directed push algorithm can reduce the packetdelay as well as the coding overhead. Meanwhile, the adoption of networkcoding technique makes the system robust to peer dynamics. Experimentalresults show that our system has demonstrated superior performance compared with previous approaches, especially under the challenging flashcrowd scenarios.