分布式存储再生码数据修复的节点选择方案

发布时间：2018-05-28 08:06

本文选题：分布式存储系统 + 计算能力异构　；参考：《计算机研究与发展》2015年S2期

【摘要】：随着海量存储数据的剧增,分布式存储系统逐渐得到广泛使用.分布式存储系统中存储节点数目的增加使得节点失效的机率增大.使用副本作为冗余的方式会带来很大的存储开销.为了减少存储开销,使用纠删码产生冗余是存储系统中一种主要的冗余策略.基于纠删码的数据修复过程中,需要在幸存节点中选择供应节点向新生节点传送数据,这就面临着修复节点的选择问题,需要设计一个节点选择机制来使得修复过程较快,提升存储系统的可靠性.结合实际数据修复过程中参与修复的节点对数据的计算速度不同,即存储节点存在计算能力异构的场景下,提出节点选择机制:星形拓扑修复算法S-SPAC和树形拓扑修复算法T-SPA-C,以加快失效节点的修复速度,从而减少整个修复过程中的修复时间.仿真结果表明,相对传统的随机节点选择策略,本文提出的选择机制可以有效减少数据修复时间.
[Abstract]:With the rapid increase of mass storage data, distributed storage system has been widely used. The increase of the number of storage nodes in distributed storage system increases the probability of node failure. Using replicas as redundant means can result in significant storage overhead. In order to reduce storage overhead, using erasure codes to generate redundancy is one of the main redundancy strategies in storage systems. In the process of data repair based on erasure code, it is necessary to select the supplier node among the surviving nodes to transfer data to the new node, which is faced with the problem of selecting the repair node, so we need to design a node selection mechanism to make the repair process faster. Improve the reliability of the storage system. Considering the different computing speed of the nodes involved in the restoration of the actual data, that is, the storage nodes have heterogeneous computing power in the scenario. The star topology repair algorithm (S-SPAC) and the tree topology repair algorithm (T-SPA-C-) are proposed to speed up the repair of the failed nodes and reduce the repair time in the whole repair process. Simulation results show that the proposed selection mechanism can effectively reduce the data repair time compared with the traditional random node selection strategy.
【作者单位】：复旦大学计算机科学技术学院;网络信息安全审计与监控教育部工程研究中心;
【基金】：国家自然科学基金项目(61571136) 上海市“科技创新行动计划”项目(14511101000) 上海浦江人才计划项目(14PJ1401400)
【分类号】：TP333

【相似文献】