Research On Search Algorithm For Small-World-like P2P Networks

P2P networks (peer-to-peer) refer to a distributed computer network in which all participants (node) share part of their own hardware resources (processing power, storage capacity, et al.), and these shared resources are provided as services to network (e.g. file sharing). The participants of P2P networks are resource providers as well as resource consumers. All nodes act as independent, interconnected participants, and their autonomous participation can greatly improve network efficiency. In P2P networks, due to the node’s lack of knowledge of global network topology, how to locate resources, reduce communication overhead between nodes and adaptive overhead have become core issues of resources search. Different topology structures use different search algorithms, and the search algorithm has a particularly important impact on the networks performance and scalability of P2P networks. Inspired by the small world model, this paper conducts in-depth research on resources search technologies.Unstructured P2P networks mainly use flooding-based search, which results in high network traffic and low efficiency of resource location. To solve these problems, this paper presents a search algorithm based on the idea of local clustering of unstructured P2P networks. Text vector is used to represent the content of the resource node. The node’s local resources are first clustered through k-means algorithm, and similar links are established among nodes according to node similarity. These links are used to improve the connectivity between nodes with similar resources. This method can promote entire clustering effect of P2P networks, improve the efficiency of resource location, and reduce network load through effectively controlling the spread of query message between similar nodes. The extensive experimental results show that compared with the Gnutella network, this method can obtain greater clustering coefficient, smaller resource query length and higher efficiency of resource location. Structured P2P networks mainly applies the DHT-based routing, which has a high efficiency of resource location, but may cause huge maintenance overhead of network topology. To solve this problem, this paper proposes a small-world-like semi-structured P2P networks. Therefore it takes the advantages of both unstructured and structured P2P networks. Drawing on the research results of the small-world theories, the method can improve the efficiency of resource location, avoid the waste of network bandwidth caused by the proliferation of flooding query messages, and reduce the cost of topology maintenance. The mathematical model and experimental results demonstrate that the method works better than Chord in terms of search success rate and average search hop.