Mémoire Du Stage Transmission Multimédia En Direct Dans Le Réseau Pair-à-pair

With the growth of the computing power of personal computer and the proliferation of broadband access to the Internet, media streaming has widely diffused. The high bandwidth required by live streaming multimedia greatly limits the number of clients that can be served by a source using unicast. IP Multicast should be a good solution to this problem, but the fact that the current Internet does not widely support IP Multicast force us to find other ways. Another existing solution is the proxy caching technique that can accomplish effective media streaming, but it cannot adapt to the variations of user locations and diverse user demands. Under this situation, the applying of P2P systems on live media streaming become a attracting direction. By using the P2P communication architecture, media streaming can be expected to smoothly react to network conditions and changes in user demands for media-streams. In recent years, there are many end host multicast solutions have been proposed, for example, PeerCast, ZigZag, Promise,Narada, Delaunay, CoopNet, and SplitStream etc. The main problem they emphasized are slight different, but all of them would have addressed the common basic streaming problem, for instance the problem of transience, the problem of control of end-to-end delay, etc. In this paper, we propose a PeerCast like P2P media streaming solution. We focused on the maintaining algorithm of a single source, receiver driven media streaming delivery tree with a desirable end host performance metrics. Our proposed solution is featured by its tree maintaining algorithm that organize the multicast tree according to the geographical characteristic of nodes. The main idea in our algorithm is the concept “nodelist” and the process of “sorting” and “adjusting”. Both the theoretical analyses and the simulation studies proved that under a client-base size of hundreds, our solution can maintain a ‘balanced” streaming tree and gain a relatively good end user performance.