DARS: A dynamic adaptive replica strategy under high load Cloud-P2P

Abstract

In Cloud-P2P system, replica strategy is utilized for obtaining low access delay and high load balance, which requires paying more attention to the time of replica creation. Most of current replica strategies usually utilize the method of afterwards adjustment fixed threshold. However, these strategies may increase a large number of overload nodes and then lead to aggregate effect of being overloaded, especially under high load condition. To deal with the problem, this paper proposes a Dynamic Adaptive Replica Strategy (DARS) based on the node’s overheating similarity. DARS addresses the problem of replica creation time, including obtaining the replica creation opportune moment and finding the optimal replica placement node, by a decentralized self-adaptive manner. Based on the overheating similarity, DARS applies fuzzy membership function to obtain the replica creation opportune moment, which enables the node to create replica before its overload. Meanwhile, DARS adopts the fuzzy clustering analysis method to find out the node with high node degree and low load which is used as optimal placement node to store replicas. Extensive experiments demonstrate that DARS obtains superior performances in access latency around 15%∼20% on average and better load balance than other similar methods under high load condition.