A Load Balance Model For Peer-to-Peer System Based On Bayesian Network

The rapid development of P2P technology during the past few years has involved more and more institutions and organizations into this research area. Various types of P2P products and services emerge in endlessly, and the application of P2P has spread over the all those regions, including file sharing, collaboration, peer-to-peer computation, real time communication, searching, network storage, intelligent agent, and Net games.Due to the randomicity of the connection, as the sharp growth of P2P network, the number of the connections between peers differ each other a lot which leads to imbalance of whole P2P network. Therefore, load balance has become one of the hottest subjects in P2P network research.Bayesian Network is a graph model which denotes the probability among variables, provides a nature way to represent causality, and predicts the future probability according to the historic information. The Bayesian Network has the features of strong probability ratiocination, context clarity, easy to understand it is one of the most effective theory and research hotspot in the area of indeterminate knowledge presentation and ratiocination.Based on thoroughly investigation and deep research on P2P technology, Bayesian Network, and load balance, this thesis starts research on load balance under P2P network, the major work includes:1. A load balance model based on Bayesian predication model under P2P environment has been proposed in this thesis. The model uses probability system load of peers predicted by Bayesian ratiocination to schedule the peer chosen algorithm, which balances the connection between the peers, and ameliorates the network system, expresses positive significant to the whole P2P environments .2.A simulated laboratory environment for P2P load balance based on gigabit switcher has been constructed. This environment includes a number of 5 server peers and 100 connections from client peers, and each connection initiate 200 requests for resources.3. Experiments for Bayesian trust network model based on none load indicator,Bayesian trust network model based on multi-layer load indicator, and dynamic loadbalance model had been taken. The results show that both the later two model canensure the load balance of the P2P network, but, compare to dynamic load balancemodel, Bayesian trust network model based on multi-layer load indicator has strongeradaptability.