The Research And Implementation Of The Indexing Technology For P2P Streaming Media System

With BitTorrent and eMule being the representatives, P2P file-sharing technology has achieved a big success in recent years. On the other hand, it is becoming more and more difficult for traditional streaming media system to meet the need in capacity. Thus, applying the P2P model in streaming media system to address the bottleneck problem becomes a current research hotspot.Resource locating is the first problem in any P2P model. Based on the background knowledge of both P2P system and streaming media system, we analyze the resource locating algorithms in depth, and comprehensively clarify the fact that the index server is the most suitable way for a P2P VoD system.Our research work is concentrated on the indexing technology for a P2P VoD system, of which the innovative work includes: (1) a fast indexing algorithm orienting high quality source peers, and (2) an indexing technology using software pipelining. These works aim at the P2P VoD system, but the technology themselves can be used in any P2P systems, or even more other fields.For the indexing algorithm, we firstly analyze the importance of high quality source peer oriented indexing. And then we propose a fast algorithm orienting high quality source peers. Finally we prove by experiments that this algorithm is able to optimize the average delay between peers in a P2P system by about 80%, and we prove by benchmark tests that the time and space cost is affordable by the service system.For the indexing service technology, we firstly analyze the flaw of current server architecture. And then we propose 2 type of software pipelining technology, one is the non-symmetric processing pipelining, and the other is zero-synchronization data pipelining. Finally, we validate the effectiveness by experiments, and we can see the performance improvements range from 5.3% to 104.9% in different configurations.In the end, we also present the collective design of a P2P VoD system, and implement the indexing server applying the technologies we have proposed.