Research Of Search Algorithms For P2P Network Systems

Peer-to-Peer(P2P) network is a new kind of complex network. It is found that even if each individual provides only a few files, enormous files can be shared by the whole group of users. In order to use these abundant resources, the key problem is to find them effectively. Therefore, the issue of search in P2P networks has become an important topic.The search in P2P networks is related to many fields, such as graph theory, statistical physics, network measuring, mathematics modeling, and algorithm design. Two main aspects of search in P2P networks have been studied in recent years: one is how to optimize and improve the P2P network models, according to the known results of the complex networks, to obtain high search efficiency and accuracy, the other is how to design more effective local search strategies based on current P2P network models.There are two kinds of P2P search algorithms: structured search algorithms and unstructured search algorithms. Although structured search algorithms can guarantee small average query hops, there are shortcomings such as difficult maintainability and nonsupport for complex search in these systems. On the other hand, unstructured search algorithms are simple, flexible and very robust, but their average query hops is much larger, and their scalability is much unfavorable.This paper studies how to overcome these disadvantages, in order to get a high level of each performance target. First, we concluded the up-to-date P2P search algorithms and their excellence and shortcoming. Second, we introduced small-world and scale-free, which are two important characters of complex networks. Then we provided two new unstructured search algorithms: diffuse search algorithm and recommend search algorithm. The former based on scale-free and small-world characters of complex networks, and follows the principle of high degree first. The latter based on self-similar small-world character of complex networks, and after the ancestor of the destination cluster is found, the algorithm will search the nodes in the cluster. Experimental results indicated that our algorithms excel other up-to-date P2P searching algorithms in many aspects.