Research On Replica Management Strategy In Cloud Storage

With the rapid increasing of data size, Cloud Storage has been widely used in huge amounts of data storage and management, and provides storage services to users. In Cloud Storage system, using replicate technology to improve system reliability, availability, and scalability are usually used. However, applying replicate also brings bandwidth overhead, data consistency, load balancing, and many other problems. In order to improve the cost benefits,storage service performance and load balancing, this paper mainly studies the replica placement strategy, specific work is as follows:On the basis of the different needs of users and node heterogeneity, this thesis introduces a new Analytic Hierarchy Process and particle swarm optimization algorithm(PSO) based replication place strategy(AHPPSO). This strategy selects CPU utilization, I/O utilization,bandwidth utilization, memory utilization, and disk space utilization as data node performance evaluation index. The improved PSO algorithm introduces weighted evaluation index. And making the PSO in the case of satisfying the constraint conditions, optimization target is optimized. The optimization goal includes the storage space utilization, the execution time and execution cost. With further consideration to reduce the storage cost, based on the analysis of the close relationship between file availability and copy number, this thesis presents a minimum replicate number strategy.Through the simulation of AHPPSO, Random strategy, Basic PSO based strategy in Cloud Storage system, the simulation results show that this proposed strategy can shorten the execution time, reduce storage cost, balance load to a certain extent, and improve the cloud storage service performance.