基于奇偶校验的三容错数据布局研究  

Research on efficient triple disk failure tolerance data placement scheme based on parity code

作　　者：刘卫平[1] 蔡皖东[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710072

出　　处：《计算机应用》2006年第4期969-973,共5页journal of Computer Applications

基　　金：航空基础科学基金资助项目(03F53031);西安市工业攻关项目(GG200312)

摘　　要：基于单容错编码的数据布局已经不能满足存储系统对可靠性越来越高的要求。对基于多容错编码的数据布局的研究受到了广泛的关注,并且出现了一些三容错的布局算法,如HDD1,HDD2等。但这些布局算法普遍存在冗余度较差、计算负载大等缺点。提出了一种基于三重奇偶校验的多容错数据布局算法TPRAID(TripleParityRAID)。该算法只需要在RAID5阵列系统中增加两个校验磁盘,通过水平、正向对角和反向对角三重奇偶校验,可容许同时发生的三个磁盘故障。该算法编码、解码简单,三重校验条纹长度相等,计算负载小,易于实现。此外,由于该算法中尽量减少了三重校验之间逻辑关联,使得该算法的小写性能比其他的三容错算法相比有了大幅度的提高。The data placement scheme based on single-erasure correcting codes can not satisfy the requirement of storage system on reliability. Now there are more and more researches on the data placement scheme based on multi-erasure correcting codes such as HDD1, HDD2. But these data placement schemes have some disadvantages such as poor redundancy, heavy computing load. A data placement scheme based on triple parity was designed, which can tolerance three disk errors. It adds two parity disks into a RAID5 disk array system, and uses the horizontal parity, diagonal parity and anti-diagonal parity to correct error. The encoding and decoding algorithm of the data placement scheme is simple, the length of parity stripe is equal in triple parity, computing load is light, and easy to implement. In addition, the performance of this scheme in small writes improves obvious comparing with other multi-erasure correcting data placement schemes, for decreasing the logical relationship among triple parity.

关 键 词：RAID5 奇偶校验 多容错编码 小数据写 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]