大规模云存储系统副本布局研究

Replica placement study in large-scale cloud storage system

在基于副本冗余机制的大规模云存储系统中,以往的副本布局算法只能部分地满足副本布局中高可靠、高可扩展并且高效的要求,针对这一问题,提出了一种基于分组和一致性哈希的副本布局算法。首先,将关联性高的存储节点进行分组;然后,通过改进的一致性哈希算法将同一数据对象的多个副本分别分配到不同分组中;最后,再通过改进的一致性哈希算法将分配到各分组的数据副本放置在组内对应的存储节点上。理论分析可知,该方法大大提高数据的可靠性。仿真结果表明,该算法能满足副本布局的均衡性、自适应性要求,并能在几十微秒内完成副本定位。

In the large-scale cloud storage system based on copy redundancy,previous layout algorithm can only partially meet the requirements of high reliability,high scalability and high efficiency in the replica layout.To solve this problem,this paper proposed a Replica Placement algorithm based on Grouping and Consistent Hashing（RPGCH）.The storage nodes were classified into different groups by their correlativity,then the replicas of one object were assigned in different groups by consistent hashing algorithm,after that each replica was placed into corresponding storage node in the group by consistent Hashing algorithm.The theoretical analysis proves that the reliability of data is improved.The simulation results show that RPGCH can assign data evenly among storage nodes and adapt well with the changing scale of cloud storage system.Moreover,RPGCH is time efficient with little memory overhead.