Research On The Search Mechanism In Structured P2p Network Based On Interest

With the development of information technology, P2P technology has received widespreadattention as a new network application model. The traditional Internet uses the centralizedtopology, there is a single point of failure, server bottlenecks and other defects, the emergence ofP2P networks can better solve these problems. In a distributed P2P network, the same status ofthe node (Peer) can communicate with each other, it can take good advantage of the edge of thenetwork resources, without defects of the centralized structure. And the capacity and resources ofthe P2P system, in theory, is the sum of all nodes in the system, therefore it has better scalability.Relative to the centralized structure, P2P technology also has the advantages of decentralized,self-organizing, etc.Nowadays, P2P technology was mainly applied in network resources (documents,computing ability, bandwidth, etc) sharing, collaboration and instant messaging, etc. However,these applications, based on P2P, depend on the routing efficiency of the P2P network in differentdegree. Therefore the key point of this research is how to improve the routing efficiencyeffectively in the distributed P2P environment. Recently, thanks to the regularity and the clarity ofits structure, the structured P2P network gradually becomes a research trend. Most researchesabout P2P technology base on the structural P2P network. Therefore, on the basis of past research,combining with the characteristics of the P2P network, this paper discusses the solutions ofimproving the routing efficiency in the DHT network. Then, the optimization scheme and theanalysis of simulation were introduced. Finally, the methods of resource location and routingmechanism in the structured P2P network were discussed.This paper mainly includes the following sections:First, it includes the analysis of the current research situation in the domestic andinternational peer-to-peer network and the introduction of its applications. Then the concept andcharacteristics of peer to peer network were described. The peer network was divided into fourcategories: the centralized P2P network, the unstructured P2P network, the structured P2Pnetwork, and the hybrid P2P network. The typical network models and search algorithms of eachcategory were discussed, and the characteristics of these models and algorithms were summarized.Through the description of the small-world model, the research center of this paper was displayed,that was, research on the search mechanism in Structured P2P Network which was based on the optimization of the nodes interest. The feasibility and effectiveness of this optimization werediscussed.Subsequently, an ontology-based P2P network resource retrieval algorithm was proposed. Inthe Structured P2P network, only exact match search which was based on keywords can be done,the complex queries and semantic queries can’t be completed. According to the above defects,this search algorithm SOC (Semantic Ontology Chord) was put forward. It can do the complexqueries and semantic queries. The algorithm was adjusted by the classical Chord model, itimproves the DHT mechanism of the Structured P2P networks. It changes <Key, Value> into<Class, Key, Value>. It makes the resource information, maintained by the original nodes, changefrom scattered to relevant. So in the query, the node will return more similar sources to improvethe precision rate and recall rate. The ontology technology was used to do the fuzzy retrieval:,nodes can not only maintain the original routing table, but also maintain the node information ofsimilar interest,.Therefore, the node can give priority to send query messages to this kind ofnodes. Because ontology can better match the resources in semantic level, it will be able to returnto more semantically similar results to further improve the recall rate. Finally, the results ofSimulation experimental which uses the PeerSim Simulator indicates that with the increase ofnodes and data, this algorithm can achieve higher recall rate than Chord.Finally, the P2P network model of a multi-interest clustering was proposed. When DHTmechanism does the node query in the structured P2P network, it cannot get the globalinformation. To improve the inquiry efficiency of DHT, reduce routing hops in the inquiryprocedures, enhance its scalability, a network model MIKAD (Multi-Interest clustering KAD),based on the interest of the node, was proposed. The model was modified from the classicalKademlia model. It calculates the multiple interests of the node through the STC (Suffix TreeClustering, the STC) algorithm. Then, the nodes join into the corresponding interest clusteringaccording to their own interest. Because of having several different interests, the node will joininto different interest clustings. However, in a certain interest clusting, other interest resourses ofthe node was invisible to the other nodes. The interest clustering, organizes, manages andmaintains itself through the super node, and was responsible for indexing the resourceinformation of this cluster. So, logically, the whole network was organized into a hierarchicalstructure, the upper layer was the connection among the super nodes; the lower layer wascommon to all networks. At last, the simulation experiment, used Peersim, shows that this modelcan achieve better search performance with the increase of nodes and data.