Abstract: The cloud storage service of open source cloud computing platform Openstack uses common hardware to build clusters of servers to provide redundancy, highly reliable, scalable data storage for the PB-level data. For the case that part nodes’ visit overheating of the cluster and another part’s resource underutilization, this paper improves the consistent hashing algorithm used in the Ring which is the most important component of the core module Swift. Experimental results show that these algorithms give consideration to load balancing to the maximum extent and improve storage resource utilization, considering cluster processing capacity of each node, while taking the current load into account and improve the overall performance of the cluster system.

Key words: Openstack, Swift, consistent hash, load balancing

摘要： 开源云计算平台Openstack的云存储服务使用普通硬件构建的服务器集群为PB级别数据提供冗余的、高可靠的、可拓展的数据存储。针对集群中一部分节点访问过热而另一部分节点资源利用不充分的情况，对核心模块Swift中最重要的组件Ring中用到的一致性哈希算法进行改进。实验结果表明，该算法能在最大程度兼顾负载均衡的同时提高存储资源的利用率，既考虑了集群中各节点的处理能力，同时也兼顾了当前负载，提高了集群系统的整体性能。

关键词: Openstack, Swift, 一致性哈希, 负载均衡