Research On Optimization Methods For Resource Search In Hybrid P2P

With the rapid growth of Internet, servers that provide all kinds of resources become performance bottleneck of the network. So the peer-to-peer network (Peer-to-Peer, P2P) is re-arousing the attention of researchers. P2P is a distributed system that can make full use of network software and hardware resources and its most important function on the Internet is to share resources. Resource search is the most basic technology achieving the function, which determines the efficiency of system. So in the P2P research field searching is a hot issue. As like other large distributed system, the performance of P2P systems depends not only on its topology structure, but also on search mechanism. As hybrid P2P system has the advantages both of centralized P2P system and of pure distributed P2P system, it becomes the current mainstream model in P2P applications on the Internet. Even though it has good search performance, it still faces a number of issues about search efficiency, such as SuperPeer (SP) overload, and inefficient search algorithm.In Application of hybrid P2P system the most important problem is how to advance search efficiency. Based on the congestion control theory and multi-objective optimization theory, this paper analyzes and researches the super-node load capacity and routing algorithms. And it brings forward several feasible approaches. The major contributions made by this paper are listed as follows:(1) In the hybrid P2P system, a SP receives the query messages and non-query messages. The efficiency of processing query messages determines the system performance, but the operation of the non-query messages will decrease the system performance. This paper proposes a new P2P topology, a Three Level Hybrid P2P System (TLHP), in which query messages and non-query messages are processed by different SPs. And then the process of dealing with query messages is divided into sub-processes to shorten the time of query messages forwarded. Simulation results show that compared with two-level hybrid P2P system, the TLHP can reduce the SPs load effectively and lower the average forwarding time of query message.(2) These methods about SP's choice share one deficiency that the SP selected by them is a single peer. If the single-peer SP fails, system performance will be greatly affected. The paper uses a Virtual SP (VSP) to act as a SP. VSP consists of multiple peers that are stronger than others in the same cluster. And it can dynamically regulate the number of its members according to its load changes. The method enhances the SP processing power and improves resource utilization.(3) In the informed search methods, the two facts are mainly considered:the sum of related resources stored in its neighbor peers and delay time between them when it forwards a message. But these methods ignore the difference among the messages with the same keyword. This paper presents a routing algorithm based on the query message status. It uses the information of message status to determine the weight of resource number and delay time. And it selects the next hop from its neighbor peers using linear weighted sum method in Multi-objective optimization theory. Finally, a large-scale network is deployed to simulate, and the results show the effectiveness and efficiency of the routing algorithm.In the hybrid P2P system, the efficiency of resource searching is determined by SPs. In order to increase the efficiency of search algorithm, reducing the SPs' load and improving the capability of SPs shorten the processing time of the query message, and using state-based query message routing algorithm improves efficiency and effectiveness of search. This paper presents three methods through experiments and simulation and proves that the hybrid P2P system can improve the search efficiency. These findings can be applied to the P2P resource sharing system to improve the performance of P2P applications.