分布式存储系统中数据修复的自适应流控机制  

Adaptive flow control mechanism for data recovery in distributed storage systems

作　　者：陈志鹏 李小勇[1] 

机构地区：[1]上海交通大学信息安全工程学院,上海200240

出　　处：《信息技术》2017年第11期139-144,149,共7页Information Technology

摘　　要：要大规模分布式存储系统中存在大量的数据迁移和数据修复任务,一方面要尽量缩短数据修复时间以降低数据丢失风险,另一方面需要避免这类任务对正常业务的服务质量造成冲击。针对该需求,文中提出一种自适应的数据修复流控机制。通过周期性的相对负载变化检测来动态调整修复流量大小,并根据数据危险级别和全局负载信息决定修复任务优先级。实验表明,该方法不仅能够提高大规模分布式存储系统的数据修复效率,避免对正常业务性能造成明显冲击,还具有自适应的优点。In large-scale distributed storage systems,there are a large number of data migration and recovery tasks. On the one hand,data recovery time should be minimized to reduce the risk of data loss.On the other hand,their impact on the quality of service of normal data access should be prevented. An adaptive flow control mechanism for data recovery is proposed for this requirement. The velocity of data recovery is dynamically adjusted after periodic detection of relative load changes,and the priority of each data recovery task is determined according to the data danger level and the global load information. The experiments show that this method not only can improve the efficiency of data recovery in large-scale distributed storage systems,but also can avoid obvious impact on the normal data access,and it has the advantages of self-adaptation.

关 键 词：分布式存储 数据修复 数据迁移 自适应 流量控制 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]