Correlation-Based Searching Algorithm In Pure P2P Network

Peer-to-Peer networks are becoming popular because while their participants downloading resources from others they also act as resource providers, which extends the amount of network resources enormously. How to find the demanded resource efficiently is one of the key factors which have impacts on the development of P2P networks and it has become a hot research topic in the field of computer science nowadays. In this thesis, the searching mechanism in decentralized and unstructured P2P networks (pure P2P networks) is analyzed and two improved algorithms: a correlation-based self-organizing searching algorithm and a selected-cast algorithm are proposed. A prototype system that implements the algorithms is presented. In the end, experimental results are given to show the advantages of the improved algorithms.In this thesis, the flooding-based searching algorithm used in Gnutella (a typical pure P2P network) is explored;Two improved algorithms are proposed to enable nodes in P2P network to change their current neighbors when searching resources so as to keep themselves near to the potential providers of resources and to choose some of their neighbors to send messages when receiving them from other nodes.The efficiency, feasibility and robusticity of the improved algorithms are discussed; differences between the algorithms in this thesis and other interest-based algorithms are presented.The working process of the prototype system which involves the algorithms are described and details in designing the structure and components of the system and definition of the communication message format are given.Details in the experiment are introduced and efficiency of two algorithms is proved by the experimental results.