Research On Adaptive Replica Distribution In Unstructured P2P Network

In recent years, with the rapid development of P2P technology, many applications based on the P2P technology have been emerging, which have greatly promoted the development of P2P networks. The file storage system, as one of the most important applications of the P2P networks, has played an important part in the Internet. It has become a hot application in Internet and occupies a larger share of network traffic. Due to the characteristic of the simple topology structure, high fault tolerance and adaptivity, the unstructured P2P network is widely used for building the distributed storage system. However, owing to the randomness of topology and the unevenness of queries in unstructured P2P network, the replicas in the different locations have different adaptive capacities. If the location of replica cannot adjust well with the query change, it will introduce many unavailable replicas, exhaust lots of storage space, and finally reduce the search performance.To solve this problem, a self-adaptive replica strategy based on query-driven is proposed in this paper. In this strategy, the nodes calculate the replicas fitness, judge the adaptive capacity of replicas and adjust the location of the replicas with poor adaptability in the network by using the "survival of the fittest" rule in the nature. The replicas with poor adaptability are placed in the frequent requesting nodes on the successful search path. This will greatly increase the adaptivity of replicas and enhance search performance.Due to the non-uniformity of the query, some critical nodes will receive too many queries that will lead to the overload of critical nodes. In order to solve this problem, we compare nodes’ loaded condition of current position and neighbour position and choose nodes with least load as the replica nodes. At the same time, we also propose a new access control strategy to prevent the access overload of the nodes. In this strategy, nodes provide the requested files to the requesters with a certain probability according to current load condition and access control rule, and this will greatly reduce nodes’ overload.The simulation results indicate that the self-adaptive replica strategy based on query-driven is able to significantly reduce query cost and number of replicas, which highly improves query efficiency and replica utilization. What’s more, the access control strategy can effectively reduce the load of nodes and improve overall performance of P2P network.