Research On Replica Organization In Unstructured P2P Network

With the development of the Internet in recent years,a large number of applications based on P2P networks emerged,which could avoided the problem of single point invalidation in traditional centralized system and provide a highly scalable solution.Highly available data storage service is a hot area of unstructured P2P research,through the distribution of files in different nodes,the file replication could ensure its availability and it is proved an effective way of improving the performance of the search.Due to the character of autonomous and scattered of the nodes in unstructured P2P network, the file replication and the replicas'consistency maintenance based on P2P networks had to face to the difficulty of high dithering and information decentralization.Based on the character of unstructured P2P networks,this paper has intensively studied the management of replica.In allusion to the character of dynamic of the node in unstructured P2P networks could affect the replication of file,a new replica management strategy is put forward in this paper. Firstly the behavior of P2P nodes is modeled from the perspective of space and time, and also the files'availability is modeled from the perspective of node's accessibility, and then the replica population organizing strategy is proposed by utilizing the population ecology model in unstructured P2P network. Since the relationship between the replica population and the node environment as well as the user needs can be adjusted automatically, it makes the replica reproduction adapt well to the dynamic nature of P2P networks. Meanwhile, the adaptive changes of replicas in quantity and distribution can be achieved by taking advantage of push and pull approaches.The simulation results indicate that the proposed replica population organizing strategy is able to adapt to the high dynamic P2P environment, effectively improve the availability of files and the retrieval efficiency, and reduce the bandwidth consumption.In allusion to the problem of redundancy update path caused by information asymmetry and dynamic nodes in consistency management of replica,we propose a minimum update path algorithm based on minimum spanning graph.Thought the clone transfer of message to solve the asymmetry of information and the cross-merge of information to realize sharing of information,then the redundant path could be reduced.In addition,unreachable update caused by invalidation of node could be avoid due to the particular path chose,using delay-wait method to combine the updates of conflicts and redundancy,the update time reduced. Simulation experiments show that the strategy could realize the consistency of replica with fewer update path and control update time effectively.