The Research And Application Of Improved PSO Algorithm In P2P Network

Node selection mechanism is the research focus and core technology of P2P network. Recent years, it has wingspread concerned. At present, in P2P network, the proportion of traffic generated by BitTorrent (BT) system is largest. BT system is one of large-scale P2P file sharing system actual deployment. More and more researchers pay close attention to it, and there are a large number of application, testing, analysis and improvement which aim at this system. Therefore, in this paper, we choose the representative P2P file sharing system-BT to explore the problem of neighbor node selection (i.e. NS). In BT network model, the hybrid structure of central tracker is used. In P2Pnetwork, one or more entities, which is or are used to oversee the peer clients adjacency in BT file transfer protocol, will guide the distribution connection or download connection between customer node and some other nodes. In this process, tracker will select a neighbor node list for each node which joining the session. However, the adjacency way of BT through random selection strategy gives a list of neighbor, typically contains40-80neighbors. Apparently, the neighbor nodes, which are generated by random manner, may be not appropriate.In view of the above situation, we adopt the double structure encoding binary particle swarm optimization algorithm (DS-BPSO) to solve the problem of neighbor nodes selection in P2P file sharing systems based on the BT, and we consider the selection of uplink neighbors and downlink neighbors from the actual ADSL connection. The characteristic is that the concept of "double structure encoded particle position" based on binary particle swarm optimization algorithm is introduced. The whole evolution process will be executed at the same time in auxiliary searching space and solution space based on the double structure encoding feature of particle (i.e. node). This makes the modified algorithm not only maintain the global convergence of the basic PSO, but also enhance the ability of particle trapped in local optimum. According to the difference degree of file fragments between selected neighbor nodes and destination node, uplink neighbor nodes and downlink neighbor nodes according with ADSL connection will be distinguished from the selected neighbor nodes, so the final neighbor nodes would be more appropriate and it could lay a good foundation for file sharing. In order to verify the performance of the proposed algorithm, we compare it with the BPSO algorithm. The experiment results indicate that DS-BPSO Algorithm usually could obtain better results than BPSO Algorithm, especially for large scale problems.