Research And Implementation Algorithm Of Node Organization And Management In P2P Streaming Media Live Broadcast System

With the rapaid development of Internet technology, the streaming media live broadcast system becomes more and more popular. As a result, Users also will be the amount of the accelerated growth, people's eyes are also getting higher and higher quality requirements. However, the streaming media live broadcast system in tradional c/s model has bottleneck in service ability, which makes it difficult to adapt to the large scale application. So, the technic of streaming media distribution based on P2P network has been proposed, which transfers the data source center from streaming media server to user network. In this model, the technology has high scalability and utility to meet service needs of large scale customers, from the nature of on the ease side of the media sources in the amount of traffic problems.The related knowledge of streaming media and basic concepts of P2P are introduced in this thesis. Streaming media in the investigation of the mainstream, based on research-based P2P live streaming system-PeerCast, its Framework, network protocols, channel organizations, relevant algorithm to conduct a detailed analysis of each module. It reducing the receiving effect (such as system relative blance, system stability and long delay) and the fact that some P2P softwares are bolcked by operaters for bandwidth reason are studied. The problem of node selection when joining the network is analyed by experiments, and a new solution of this selection is proposed, which builds a balance selection model of depth and routing based on the least total delay and occupied bandwidth when sigle node joins. The nodes with small depth are selected in this model for less delay. In addition, the nodes nearby are selected for reducing the repeating lines occupation. The situations of absolute balanced tree and linear tree are avoided in this model, and the bandwith occupation of operator network is saved. Besides, an improve solution when node leaves is proposed. When a node leaves the network, all of nodes of the disconnected sub-tree would try to access the same node, which would heightened the pressure of this accessed node. So, a selection model of uniform probability distribution based on the number of relay nodes is build for mutiple nodes joining when single node leaving. In this model, the larger the relay node amount a node has, the bigger the probability that this node would be selected as the parent node is. The multiple nodes of the disconnected sub-tree can randomly and respectively connnect their parent nodes, which would change the situation of big node burden, network congestion and low joining delay, compared with the way of accessing a single node. And then the related verification is presented by experiments, which shows the performance of the live broadcast system improves obviously.Finally, a conclusion of the research work on PeerCast streaming media live broadcast system and the related work in the future are stated.