云存储环境下副本选择策略研究

Study on Strategy of Replica Selection in Cloud Storage Environment

云存储服务提供商为了满足各类云用户的存储需求,一般采用划分固定大小的数据块、冗余备份等技术来存储数据,关于块放置、最佳副本选择、副本粒度等存储机制的研究一直是加快大文件存取速度的重要内容。面向云存储系统中存储节点的异构性,设计了一种采用层次分析法对节点性能指标加权并依据加权指标改进粒子群算法的策略(AHPPSO)。通过引入与存储节点性能相关的加权评价矩阵,使得粒子群算法向综合性能较高的节点进化,在不增加存储空间成本的基础上,加快了存取数据的速度。在自主搭建的云存储系统中实现了该策略,实验结果显示该策略能够适应多种用户需求,并且在一定程度上实现系统负载均衡。

In order to meet the needs of various users for cloud storage,cloud storage service providers generally divide the data into fixed size block and use redundancy backup technology to store data. So the researches on storage mechanism of the block placement, the best replica selection and the replica size have always been hot spots in improving the transmission speed of big file. According to the heterogeneity of storage nodes in cloud storage system, the improved strategy which uses AHP to weight the index of node performance, and uses the weighted index to improve particle swarm optimization algorithm（AHPPSO） was proposed. By introducing the weighted evaluation matrix associated with the performance of the storage node, PSO evolves toward the node of high comprehensive performance, improving the data transmission speed without increasing the cost of storage space. The strategy was realized in a self-built cloud storage system, and the experiment result shows that this strategy can adapt to the various needs of users, and achieve the system load balancing to a certain extent.