Research On Self-adaptive P2P Network Model Based On Peer's Serve Capabilities

Peer-to-peer network P2P is currently popular technology in the field of computer network as a new network model. It has some advantages in high search efficiency and good scalability etc, effective solution for the C/S mode single point of failure problem. Thus, many of P2P applications are rapidly and widely used in many fields, such as e-commerce, file sharing, instant messaging etc. As P2P application is built based on the P2P model, so how to build a reasonable model for P2P has become an important issue.The study of P2P network model has been a hot and difficult topic in P2P research field. Current P2P models have their flaws, centralized P2P model has single point of failure problem; Fully distributed unstructured P2P model has a poor scalability; Fully distributed structured P2P model needs a complex maintenance mechanism and its dynamic adaptability of the network is very poor; Semi-distributed P2P model is combined with the advantages of aforementioned several models, but there are also some of inherent limitations. In this paper, we have analyzed the heterogeneous of nodes in P2P network and the dynamic of network, proposed BSC-P2P, which is a self-adaptive P2P model based on peer's service capabilities. The model can adjust connections between nodes according to the node's heterogeneity definition of node service capabilities, self-organizing formation services group, which can improve its adaptive capacity. Moreover we in-depth study the issues of super node selection model, and present a new method for super node selection based on CPU dynamically processing power. The model can predict the dynamic load of CPU using on RBF (Radial Basis Function) neural network, and assess the dynamic processing power of super node's CPU by the dynamic loading of CPU, which enables the ordinary node to select a super node of strongest handling capacity to provide its service.In order to verify the performance of BSC-P2P model, we have completed simulation experiments and compared the performance between BSC-P2P model and Gnutella 0.6 model from four aspects. The results show that BSC-P2P model is more effective than Gnutella 0.6 model.