Research On Classification Based Retrieval Mechanism In P2P Video Sharing System

Some new applications of internet such as Video Sharing system adopt a traditional Client/Server structure, which can not meet people's growing demand. The main reasons lie in the high cost and work-load of servers. All these reasons make it hard to improve customers' watching experience. some of the characteristic of this Video Sharing system ,including short length of video files, short time of playing the videos, large number of the videoes, large number of users, instantly request instantly watch, etc, make it hard to apply the existing matured techniques in the peer-to-peer field to Video Sharing system.In the analysis and study of all kinds of Peer-to-Peer network file-sharing system and streaming media system. We raised a scheme in peer to peer network video sharing system ViviYard which applied the Peer-to-Peer technology to network video sharing programs. In order to improve the user experience, the problem that how to retrieval peers storing a certain video file has become a core issue. In the analysis and study of all kinds of Peer-to-Peer networks retrieval technology, the ViviYard system presents a retrieval mechanism base on classification, which improves retrieval speed greatly.This retrieval mechanism modifies current retrieval mechanism in p2p networks from the following four aspects. First, all the nodes in the system is divided into ordinary nodes and super nodes, which take the responsibility of indexing the information. Second all the videos in the system are classified into various categories; all the super nods are deivided into groups by the category of videos. Then the super nodes in the same group establish neighbour relationships and exchange index information periodically with each other. Finally, super nodes maintain the three-tier index information f for the consideration of cache improvements. There are two core algorithms in this retrieval mechanism, one is the super node neighbor selection algorithm and the other is index information exchange algorithm. Super node neighbor selection algorithm selects the appropriate nodes as a it's neighbor nodes, index information exchange algorithm exchange of information between the neighbors through Gossip-proliferation.The retrieval mechanism has been designed and implemented in ViviYard, the P2P sharing system. However, due to the limited number of users at present, access to the data and logs are also limited either. Therefore, much work focus on the simulation to testify the algorithm of rapid retrieval by classification. From the result, it is proved that the retrieval mechanism is better than traditional retrieval mechanism.