Study Of Models Of P2P Networks

According to recent years' researches, we found that many real systems can be described with a complex network. These complex networks have some common characteristics, such as small average path length, big clustering coefficient, power-law degree distribution, small world and scale-free etc, these characteristics are the results of gradual evolution for the complex networks to accomplish some special functionality. Complex networks became an important way to study complex systems, and furthermore, the models of complex networks are important tools to study the complex networks.P2P is a distributed network, it is the hot point of P2P network. To establish a model which can perfectly describe the really networks is significant helpful to the research on the network's structure and behavior. Nowadays, in order to present an available and feasible scheme for P2P system, the study on P2P network model mainly focus on the routing strategies, aimed at scalability, load balancing, as well as the management. But it cannot describe the P2P network. In this thesis, a new model about P2P network is presented:Regarding of some limitation of existed network models to model the realistic world, some extending its algorithm, a new network model is be formed in the thesis, according to the characteristics of group growing and choosing the best from the part of the network(the global optimum in some large-scale network is unrealistic and unfeasible, in general, the new node may not have the ability to fully understand the situation of the entire network, or the cost to get the overall information is too high, so it does not need to connect to the largest degree of the original network nodes) ,and the attractiveness factor of nodes(In the real system, the new node connect to the node with a larger degree or a node with high an attractiveness factor. Attractiveness factor is also known as the attractiveness of the node itself). After a comparison of their behavior base on the parameters of the static statistics(dgree distribution,average path length,clustering coefficient),it can be easily find that the topological structure and development of the new model is more close to p2p networks.