对象存储系统中基于负载均衡的设备选择算法

为解决大规模海量存储系统中用户数据访问过程中出现的系统负载不均衡现象,深入分析了对象存储节点的负载特征,提出一种对象存储设备的选择算法,检测和计算各个对象存储设备的实时负载信息,并依此从中选择当前负载较低的设备进行数据访问.测试和仿真结果均表明:对于I/O请求中写请求占主导的负载,该算法实现了负载的均衡分布,消除了单点过载,缩短了系统响应时间.该算法不仅适用于写请求到达时的设备选择,也适用于重要数据的备份选择以及链式MDS中后继节点的选择,具有较广的应用范围.

Load balance based device select arithmetic in object-based storage system

In order to resolve the load imbalance caused by user data accessing in mass storage systems,analysed the load characteristic of the object-based storage devices,proposed an device select arithmetic,which can detect and account the real time load information of object-based storage devices,then decide those devices being used in the data accessing.Both test and simulation results show that during the written primary I/O load,this algorithm would make the system work load distributes uniform among storage nodes,eliminate over-loaded nodes,and shorten the response time.Besides the usage in the devices choosing for write requests,the algorithm has a wide usage,such as in backup device choosing for important data,in the subsequent matadata service node selecting for high reliability.