一种全分布式的数据迁移算法DBDS

摘要

Data migration is a significant method for maintaining the load balance for large scale storage systems. It can greatly improve data access performance. However, most existing data migration mechanisms rely on a centralized algorithm, which seriously restricts the expandability of the system. The authors make efforts to devise a fully distributed solution called DBDS ( Dimension by Dimension Spreading) to solve the data migration problem. Simulation results show DBDS enjoys both load global balance and consumption decrease during migration, so that it adapts to large scale distributed storage systems.%数据迁移是大规模存储系统实现负载平衡的重要手段,可有效提高数据访问的性能.然而,现有数据迁移机制大多依赖子集中式算法,存在扩展性方面的不足.提出一种全分布式的数据迁移算法DBDS(Dimension By Dimension Spreading)来解决系统扩展性方面的问题.仿真实验表明,该算法在保证负载全局均衡性的同时,可有效降低迁移过程的消耗,适用于大规模分布式存储系统.