Study On Topology Optimization Of Unstructured Peer-to-Peer Overlay Network On The Base Of Link Modification

In recent years, as the networks technology advance and the power of PCs enhance, P2P networks rise rapidly. P2P networks change the traditional C/S paradigm in which the servers are in central positions in networks. Each peer in P2P networks has equal functionalities and responsibilities: each peer can act both as server to supply resource, and as client to utilize resource provided by other peers. This change makes P2P networks avoid the bottleneck caused by central servers in C/S paradigm, and can make use of the resources in the networks more efficiently. Furthermore P2P networks are scalable,robust,and can be easily deployed. For these reasons, P2P applications prevail over Internet, and are widely used.Resource location is an important and basic service in large-scale P2P networks.The performance of resource location influences the netwoks'availability greatly. All the applications of P2P networks run on the top of overlay netwoks. The topology characters of overlay networks are important factors for resource locating. And thus topology optimization becomes an approach to improve the performance of resource location. In this paper, we study the topology optimization question in unstructured P2P networks. The major contributions are as follow:We propose a method for overlay network topology optimizing which contains two stages. This method can both improve the search performance and the connectivity of unstructured P2P network. In the first stage, each node search for nodes which have similar intrests in the network, and sets links to them. Thus, the nodes have similar intrests gather togother by changing links in the overlay network. In the second stage, we focus on the edge connectivity problem. The operation in the first stage may cause cut edges in network, methods for discovering and elimating these edgs are proposed. The experiment results show that this method for overlay network topology optimizing can improve the query success ratio, reduce the traffic cost, and thus make networks more scalable.The topology of Unstructured P2P network is loose, and the resource distributes random in the network. It makes difficult to make balance between traffic cost and success ratio in resource location. To solve this problem, we propose a topology optimizing algorithm based on preterit query information and converse-select mechanism. Nodes in the network trace the the queries answer information of other nodes, at the same time, based on its own interest, processes the converse-select mechanism. The nodes gradually establish high efficient links between nodes which have similar interests. This method can improve the search performance and restrain"free-riders".The connectivity of networks is the foundation for nodes'communicating. P2P networks are highly dynamic, nodes and the links between them change frequently. This causes the fall of nodes and edges in networks which may be harmful to the connectivity of networks. Besides the process of topology optimization often involves the modification of edges, and usually the influence to network's connectivity is ignored. In fact, the edges in network are important to network's connectivity. We define the"cut edge"in P2P environment, and propose distributed method for finding and eliminate them. Experiment shows the efficiency and the availability of the method.