Paired Caching Approach To Solve Churn Problem In P2P Networks

Churn is the essential characteristic of P2P network. Its impact on P2P network is negative in most cases.The network structure may be disrupted in high Churn, a large number of invalid resource indices may appear in a short time in this case, so that resource retrieval and node management would be inconvenient. To eliminate Churn effectively become the basis of running a P2P application efficiently.There have been many studies on the Churn problem, each have a different focus, the main research interests on Churn are the life time of network peers, mathematical modeling on Churn and solutions to Churn. The approaches of resisting high Churn in relevant researches could be divide into several categories,such as file redundancy, segmentation and redundancy, routing selection, changing the existing topology, etc..However, no universal solution has presented.This paper summarizes the Churn problem systematically, including the definition, digital statistics, mathematical modeling, its impact on various P2P networks and the existing solutions to Churn. By summing up relevant papers, a generic solution to Churn is presented. The lifetime of node in P2P networks is divided into several phases,the node will be in the latency period before it join the P2P network. In order to control all the peers more effectively, P2P application is divided into several modules based on their functions, make use of the complementarity between the exit peers and the join peers, a network structure, called Constellation, to maintaining a stable network resource in resource maintenance module, was constructed. Several peers in Constellation form a group which called Constellation group. In Constellation group, two nodes compose a pair. In the pair, the two cache each other's resources in their lifetime, respond to resource requests of its spouse when the spouse exits, so that the influences to P2P system are controlled. Constellation group is establishing distance-based, so it has good locality. Bandwidth consumption and computing time needed by the maintenance of Constellation is low. Finally,to verify the effectiveness of this approach, the paper gives a formula to Churn frequency.Simulation was taken by programming an application of a collaborative computing system, experimental results show that our approach has good ability on Churn resistance, also it has good feasibility and controllability.