Research And Application On Sector Error Coding In The Disk Array

With the development of science,data are growing at the unimaginable speed.Lots of data bring the severe challenge to the storage system,so reliability of storage is very important.In the existing storage system,application of disk array is extremely extensive.Disk array has the disk errors,latent sector errors and undetectable errors.In this paper,the author focuses on the sector error codes.In the existing sector error codes,majorities of them seldom repair the sector errors.Li et al proposed STAIR,which can repair more sector errors,but it is only limited to the specific sector error mode.Bian et al proposed a special sector distribution p for verification.The distribution in STAIR codes is used for coding to solve a problem of specific repair,but just increases storage overhead.The STAIR codes are obtained by column MDS code and row MDS code.The finite field is determined by the code length of two MDS codes.On the basis of verifying the sector distribution p,MDS code length used by the STAIR codes is reduced to obtain two improved STAIR codes,thus the finite field is slightly reduced in certain range.Through the comparison of performance,the calculation complexity of the improved Up-STAIR code is larger than Up-STAIR code,showing that the performance is sacrificed to obtain decrease of the finite field.The computation complexity of the improved Down-STAIR code is smaller than Down-STAIR code,thus it is a good improved method.Generalized Concatenated Types of Codes(GCC)proposed by Blaum et al can repair more sector errors.The shortcomings are the same with the STAIR codes,but the finite field is smaller.In order to conveniently make a comparison with the improved STAIR codes,the author in this paper elaborates the encoding process of GCC just like the improved Down-STAIR code.Both of them use the same row MDS code,but column code of GCC are not MDS code.Moreover,the finite field of the line codes is only limited to the disk number n.Therefore,the finite field of GCC is smaller than Down-STAIR code.The computation complexity of GCC is larger than the improved Down-STAIR code,so as to reduce the finite field.The rack construction is the storage system of which topological structure is similar to the disk array,thus the error codes of the sector can be used in the rack construction,so as to maximize coding.As application,it is necessary to notice transmission bandwidth of transmission data between racks.In this paper,the author applies the improved Up-STAIR code and GCC in the rack construction and adds locality to each rack,thus transmission bandwidth between racks is reduced.Through performance comparison,it can be observed that the improved Up-STAIR code in the rack construction is the same with transmission bandwidth of GCC,but the computation complexity can't be directly compared,because computation complexity of a code in the two codes used by both of them is larger.