Study On P2P Network Search Algorithms Based On Classification And Feedback

Along with various resources like numeric terminal, servers andnetwork bandwidth continually maintain Moore's Law, the more directsharing way will make the communication efficiency and bring a newdevelopmental tide for the information society. P2P (Peer to Peer)network is the main candidate. Nodes may connect not through servers,but directly share information resources, the processor resources, memoryresources, and high speed buffer resources. Peers can obtain theinformation from other servers, simultaneously also can respond othercustomers as the server. Therefore, it utilizes distribute resources, reducesthe fixed equipment expenses which the central networks need.At present, there are two kinds of interest-based classified searchalgorithms in P2P network, one classifies in advance, it has unitarypattern and bad dynamic peers. The other uses clusters according to theinquiry news, the cluster is indefinite and the speed is slow. Thesemethods don't consider communication delay and the messages which arenot belong to node's interest. This article proposes a P2P network classification search algorithm based on P-paradigm model. It uses P-paradigm model to compute the similarity of interest, it consideres thecommunication delay to establish link, and partitions similar nodes. Theinquiry node sends messages to the central node, and central node sendsthem to other central nodes. After the central node receives the inquiryrequest, if the search subject arranges at this group of attention front K (Ktakes generally 1～3)position, it searches all nodes in thisgroup. Comparing to MSW(C) searching algorithms, the number ofmessages reduces 50.75%, the time of discovering the first documentreduces 15.19%. Comparting to MSW(H), when the quary is the intrest ofthe node, the time of when this algorithm discovers the first document isclose, the ratio of finding files increases 1.36%; when the quary is not theintrest of the node, the ratio of finding files increases 123.21%, the timeof when this algorithm discovers the first document reduces 71.66%. Thetheoretical analysis and experiment results show that this algorithm issuperior to the MSW searching algorithms.At present, there are two kinds of feedback-based search algorithmsin P2P network. One returns some popular or unpopular resources, makesthese resources well-known, the other returns the information of neighbornodes. But these algorithms don't consider the content of nodes, thehistorical messages, historical providing rate. This article proposes a P2P network feedback-based selecting algorithm. It uses the content of nodesto computate advantage of nodes. It uses average shoot ratio and similarybetween the history queries and the new one to calculate searchingsucceed. It dynamicly adjusts neighbor nodes, enhances the P2P networksearch stability, and adjustsαto calculate the dynamic advantage.Comparing to NS selecting algorithms, the number of messages reduces11.84%, the time when the algorithm discovers the first document reduces1.61%, the ratio of finding files increases 17.50%. The theoreticalanalysis and experiment results show that this algorithm is superior to theNS selecting algorithms.