Reseach On P2P Network Based Media Streaming System

With the Internet having entered the age of multimedia communication, media streaming, as a technology for multimedia transmission, is becoming the mainstream application over the Internet. However the traditional client/server(C/S) based media streaming system has a capacity bottleneck therefore is not suitable for large-scale media streaming applications. For this reason, P2P based media streaming systems are proposed by researchers, in which playing nodes cooperate in transferring media date. With the contribution of ordinary nodes, the overall system capacity is scalable and will increase with the amount of playing nodes.In this paper, we propose a P2P media streaming system called ShareWatching, in which nodes cache and share with each other a portion of the media data consumed so the load on the media server is alleviated and more nodes could be served. According to a self-organizing algorithm, nodes playing the same media file will automatically form an hierarchical tree-like P2P network, rooted at an index server. A resource reporting mechanism is applied on top of the hierarchical tree to make resource information gather at the index server which is responsible for answering queries. This makes the proposed system have high searching efficiency and be able to balance the load on nodes with different abilities. When selecting which nodes to getting data from, ShareWatching's sender selection policy takes into account bandwidth, hop distance and alive connections of the candidate sender , so it can make best use of playing nodes to make load on the server as little as possible. It can also lessen the burden of backbone network by avoiding long-distance data transfer. During data transmission, the dynamic monitoring and controlling scheme of the system can ensure the download speed is fast enough for playing the media. Furthermore, ShareWatching's reputation mechanism helps to encourage sharing to improve system performance.Through intensive simulation we demonstrate that with each node caching and sharing 5% or more of the media file, the proposed system can greatly outperform the C/S media streaming system with respect to system capacity.