Optimizing Technique Of P2P Overlay Network Based On Super-peer Architecture

Rapid growth of P2P applications has attracted worldwide attention. P2P computing has become research hotspots in industrial and academic world. P2P network is a logic overlay network building on the existing physical network. And the super-peer overlay network is a higher level logic overlay network building on P2P network.The variou functions of the P2P system are completed by nodes communicating and collaborating with each other through all these overlay networks. At present the efficiency of construction protocol of super-peer overlay network need to be improved. A large number of semantic information contained in nodes has not been fully utilized. In addition the topology mismatch problems and free-riding phenomenon are popular in P2P overlay network. All these situations lead to the growth of delay, the degradation of searching performance, and hindered the development of P2P technology. So how to optimize the overlay network of P2P, how to reduce network bandwidth caused by the search mechanism, how to reduce the delay of resource searching, how to provide better performance, and at same time to ensure robustness and scalability are still an important research topic in P2P systems.This paper investigates the optimizing technique of overlay network in P2P network based on super-peer architecture. The main research work of this paper is as follows:(1) To improve the insufficiency of super-peer overlay network, the semantic information is introduced into the process of building P2P overlay network. And a P2P searching model, SSP2P, is put forward. The nodes in SSP2P are organized into varied area autonomy system (AAS) according to their physical topology and small world theory. Owing to heterogeneity the node with highest capability is selected as super-peer in each AAS. And super-peers are formed into many super-peer semantic overlay networks (SSN) based on semantic relationship. Thus a hierarchical P2P network is formed. The organization of the node, the construction of SSN, and the search mechanism in SSP2P are also discussed. Through the simulation the rationality and validity of model are proved. It has a higher recall, less messages, smaller response time, better robustness and better expansibility than that in Lime Wire.(2) To improve the efficiency and the topology mismatch problem of P2P overlay network, a new method of topology construction, CLSP2P, was introduced. Not only the physical location of nodes, but also the semantic information of nodes is considered in the topology construction of CLSP2P. Nodes in CLSP2P are first organized as autonomous domain based on their physical location. Then the nodes in a domain are formed into different semantic clustering according to their resource shared. The CLSP2P is supervised by domain and clustering simultaneously. Simulation results show that the query delay and traffic redundant is effective reduced in CLSP2P, at same time the query success rate is ensured.(3) Aimed at free-riding problem in P2P network, using game theory as a tool, a new incentive mechanism based on supervision game is brought forward. Through the introduction of the supervision mechanism, nodes must choose the rational game strategies:if node uncooperation it would be punished, and if node cooperation it would get benefits. And the mixed Nash equilibrium strategy adopting supervision game is figured out. The game between supervisory cost and participating income is analysed too. By adopting suitable supervisory probability, the nodes in our incentive mechanism will actively contribute their resources, and the cost is small at the same time in simulations. It improves the performance of P2P network.(4)Differing from the general super-peer selection which is only based on physical capability and unconsidered free-riding, a super-peer selection mechanism with incentive, SSMI, is presented. Any node in SSMI has a value called utility. The utility integrates the features including physical capability, voluntary and anticipation to future. The super-peers are selected based on the utility value. At same time super-peers are granted with supervising privilege and they can identify and prevent the free-riders through the feature of voluntary. It's been shown that SSMI increases the total number of file shared, improves the file querying success rate, decreases the average querying hops, and ensures the cost is small at the same time in simulations. The super-peer overlay of SSMI is steady and robust.