Response Time-based Replica Managemnet Strategy For Cloud Storage

With the rapid development of the information technology, the scale of data isgrowing exponentially. In order to store and manage vast amounts of data, cloud storagesystem emerges to effectively organize extensive storage resources. As one of theperformance enhancing techniques, replication technique has been widely used in cloudstorage systems to increase the data availability and the system performance. Lacking of asingle request in the system, most of existing cloud storage sytems increase the responsetime of user requests.By taking the response time in the storage network into consideration, this paperpresents a response time based replica management strategy refferd to RTRM strategy,through making full use of the topology of the storage network in cloud systems, RTRMstrategy want to reduce the response time of each single request. RTRM strategy consistsof a replica placement method based on the response time and the storage capacibility, areplica selection method based on the service capacibility, a dynamic replica createmethod based on the user access frequency and a replica delete method based on the datareliability.Through the simulation of response time based replica management strategy inOptorSim, the results show that the average job time of response time based replicamanagement strategy is32%lower than the built-in replica management strategies inOptorSim, the network utilization is48%lower than the built-in replica managementstrategies in OptorSim and the number of total replicas is16%lower than the built-inreplica management strategies. It meas that the response time based replica managementperforms better in reduce the average job time and the network overhead, also the numberof replicas.