一种改进的云存储系统容错机制

Improved fault-tolerance mechanism in cloud storage system

在云存储平台下,提出了一种基于访问统计的自适应容错机制SFMAF,该机制通过近似最近最少使用算法维护一张文件访问频率表来自适应调整容错方式。SFMAF对于常读取的文件采用副本冗余机制;对于不常读取的文件采用Reed-Solomon(RS)纠删码容错机制。实验结果表明,SFMAF相对于副本冗余机制,在CPU和内存使用率可接受范围内的增加上,减少了系统内部数据的传输流量,即减少了系统存储空间。

This paper presented a self-adaptive fauh-tolerance mechanism based access frequency （SFMAF） in cloud storage platform. This mechanism was implemented by maintaining a file access frequency table using approximate least recently used algorithm. If the file was read frequently, it used replica redundancy mechanism. If the file was seldom read, it used RS erasure code mechanism. Experimental results show that, comparing to replica redundancy mechanism, SFMAF increases CPU and memory utilization within the acceptable range, and reduces the data transmission flow within the system, that reduces the storage space in system.