Research Of Caching And Scheduling Mechanism In Streaming Media System Based On Bittorrent

BitTorrent is the most popular software of P2P, for its characteristics of large-scalecontent distribution and rapid downloading. BitTorrent system has a lot of similarities withmesh topological structure of P2P streaming media system. But the content-distributionmechanisms will lead to download media pieces out of order and it is not suitable fortime-sensitive multimedia streaming. In view of high efficiency and open source ofBitTorrent, node scheduling mechanism of BitTorrent is improved and integrated in themultimedia streaming. Simultaneously the media server is introduced into the system, andthe caching and scheduling mechanism of the media server is designed. On the basis ofcomprehensive analysis for the current research in this area, this paper has further deepresearch on the problems of the caching and scheduling mechanism of the system ofstreaming media. The material contents are as follows.First, the dissertation comes up with the node scheduling strategy. Nodes use slidingwindow mechanism and distinguish priority of the media segments, meanwhile theprefetching strategy is introduced. In addition, the sliding window size of node isoptimized based on the characteristic of the node. Finally, some best neighbor nodes areselected to serve the node.Secondly, caching and responding request scheduling mechanism is proposed basedon media server. Media server is added as the supplement of the node, the dissertationresearch the caching algorithm of the media server. First, the cache replacement algorithmCAVF (Caching Algorithm based on Value Function) is based on the value function, whichtakes the popularity and the relation between supply and request of media segment intoconsideration. Secondly, on the basis of caching algorithm, according to the characteristicsof the request data segment and the band of the node, a responding request schedulingalgorithm based on segment priority is designed.Finally, the caching and scheduling algorithms proposed in this dissertation aresimulated on the GPS. It compares with other typical algorithms from continuousplayback、start up delay、cache hit and server’s load respectively, and forecasts next step of work.