Cluster Cache Based Peer Selection Strategy And Its Application In P2P Streaming Media System

With the increasing popularity of internet and the rapid development of transmission technology, internet steps into the time of multimedia communications. Faced with limited network bandwidth, streaming media technology provides the conditions to achieve real-time audio and video transmission,and P2P technology based streaming media system has been widely used for its good scalability.The playback of streaming media needs high bandwidth, good resources and a stable network environment, but in the P2P networks, the performance of nodes varies, such as bandwidth, computing power, storage capacity, geographic location, etc., which result in various service capacity to others,also,a node in P2P networks may exit or fail at any time, the instability of nodes will lead to the unstable service, so how to build a stable network environment to ensure quality of service is the most important issues facing the P2P streaming media system.Combined with the characteristics of streaming media, this paper first analyzes two main areas about streaming media transmission quality of service: the client start-up delay and playback jitter, and then corresponding node cache strategy and the peer selection strategy is proposed.In P2P streaming media system, the factors that impact client start-up delay mainly refer to resource query delay and streaming media transmission delay. In order to reduce the resource query delay, the requester needs to find resources as soon as possible, and in order to reduce the transmission delay, it is a need for resources to closer to those requesting it, in other words caching resource in client-side to reduce the start-up delay. On the other hand, the user's demand is often related to each other in time. when some hit shows, it will attract a large number of users who will be potential providers of resources, so the resources in this period can be utilized to serve for future, which will increase capacity of the entire system. In addition, by the small-world model, the user's data-sharing is in line with the principle of data locality. Users within a region download files from other regions, which may be used by other users within the same region. Therefore, grouping the physical neighboring nodes into a local network can effectively use the shared data, thus improving other user's quality of service the local network.From the above analysis, a cluster cache based peer selection strategy is presented, first of all, a two-tier network of DHT is built, the upper layer is used for high-speed routing, and the lower is self-organizing cluster, through caching resource query records and some of the media block in cluster, start-up delay is reduced, then based on bandwidth priority as well as node life, resource peers are selected to decline start-up delay and playback jitter rateIn order to evaluate the proposed strategy, based on Kademlia protocol streaming media on-demand modules are design, and compare our strategy with bandwidth first strategy based on Plethora. Through caching peers'lookup records and part of media blocks, then stable resource peers are selected, so start-up delay and playback jitter rate are declined.The main innovations in this article are as follows:1. A cluster of double-buffering mechanism is proposed, through the cluster nodes caching query records and some of the media block, the resource query delay and transmission delay are reduced.2. In accordance with the order of the distance, bandwidth, node life to choose resource nodes,peer selection strategy is raised to reduce playback jitter.