P2p Network Search Algorithm Research

Resource searching is one of the core problems in the P2P system. Searching efficiency is the key of the system's capability and also the bottleneck of the P2P development. In view of this, scientists reserached various efficient algorithms, including the Gnutella system based on unstructured random topology network and other algorithms based on structured DHT, such as Chord, Pastry, CAN, and Tapestry.Based on the research of Chord algorithms, we put forword the JChord system, which improves the resource searching strategy from three aspects. First, according to the nodes' comprehensive differences in performance and their actual physics IP addresses, we select out super-node, which divides the network into the main net and the sub net. The main net is constituted by super-nodes which has the character of self-adaptive, and ordinary nodes compose the sub net which makes the super node as a center server, at the same time we get the super-node's ID from its actual physics IP address, which matches the physical topology and logical topology; second, remove the redundant information in the finger table, then add the same amount of useful information, which expands one hop's extent from semi-ring to the whole ring; third, we add a cache for every super-node, which save the resources indexes searched recently, for some hot resources, previous query results can be fully used.In order to verify the performance of the new JChord algorithm does bring better performance than the original Chord, we first prove it from theory, then write a program on P2PSim for simulation, the result shows the theory analysis is right.