Coding Algorithm Based On Repeat-Code And Binary-Matrix In RAID

With the development of internet, network storage is becoming more and morepopular. The importance of the data value makes most of the current network storagesystems use RAID fault tolerance technology to protect the reliability of data.However, as the access load of web server is increasing rapidly, it costs time toundertake encoding and decoding operation in RAID when data storing or recovering,and this leads to the current network storage systems face the problem of accessperformances. However, many of the existing coding algorithms focus too much onthe fault recovery capacity but less on the performance of coding and decoding. As aresult, it leads to take large overhead of time,and restrict the access performance ofthe server in some extent. So it is necessary to improve the efficiency of encoding anddecoding methods. Our main works can be summarized as follows:Firstly, this paper provide a code optimization scheme based on repeat-code forthe problem that many repeat operations led to low efficiency of in array codes whenencoding and decoding. First, this scheme looks for the items which need to computerepeatedly and computes the result of these items.Then these results are used as awhole input data called repeat-code to reducing the number of coding operations. Inthe scheme, the way of looking for repeat operation items is using edge coloringalgorithm of graph theory, and this algorithm can find the most repeated operation,thereby to minimize the number of operations.Through comparatively analysing, theencoding and decoding performance of this optimization scheme is obviously betterthan the original scheme in the small fault-tolerant environment.Secondly, it is hard to achieve the traditional RS code as vandermonde matrixrelates to the multiplication and division multiplication and division in finite fieldswhich often leads to large computational complexity and implementation difficulty.To solve this problem, this paper provides a method for coding RS code based onconstructing a binary matrix. This method is based on the concept of isomorphism infinite fields, using binary matrix elements to replace the vandermonde matriceelements. So there is only XOR operations in the encoding and decoding process.Simultaneously, according to the characteristics of the former w columns in the binarycoding matrix, the paper propose a improvement based on multisection: the datas offirst redundancy disk is divided into w copies, and then use the w copies to generate the other datas, so it can reduce the numbers of the XOR operations. Analysis showsthat the improved RS code is not only easy to implement but also having highefficiency.