Research On Reconstruction Technology Of Fault Tolerant Storage System Based On Erasure Code

With the dramatic increase in stored data,data failures occur frequently in the system.The existing fault tolerance methods mainly include replication and erasure coding.Erasure code is an important fault tolerance strategy in storage systems because it has less storage overhead than replication.Data reconstruction is the focus of erasure coding,and its research on reconstruction technology is mainly carried out from two aspects: encoding scheme and reconstruction process.Block codes in erasure code reduce the amount of data read during data reconstruction by grouping data blocks.The block codes encoding structure is simple,but the existing block codes have poor correlations between them,which will lead to problems such as reduced fault tolerance.At present,there is less research on the block codes reconstruction process,and it is difficult to reconstruct the block codes Combined with coding scheme.In view of the above problems,this paper will improve the reconstruction technology of erasure code from two aspects of encoding scheme and reconstruction process to improve the reconstruction efficiency of data in fault-tolerant storage systems.The research work of the thesis is divided into the following two parts:(1)Aiming at the problem of low fault tolerance caused by poor association between multiple groups of block codes,this paper proposes JLRC based on the idea of block codes.Data are grouped and a single block of data is selected in each group,and a check block is generated to jump to strengthen the connection between the groups.Then the coding structure and coding algorithm of JLRC are given,and the main process of JLRC reconstruction is proposed and analyzed.By adjusting the parameters of the JLRC itself to compare the changes in reconstruction performance,and comparing experiments with other commonly used types of erasure code,the experimental results show that the JLRC can improve fault tolerance,fault tolerance and reduce the cost of reconstruction.(2)Based on the above-mentioned skip erasure code JLRC,a tree-structured reconstruction based on bandwidth topology and computing capabilit y called TRBCC is proposed.First,a weighted undirected graph between the newly born node and the participating reconstruction nodes is generated by using the bandwidth between the nodes as a weight,and then the maximum reconstruction tree is constructed according to the TRBCC algorithm,and it is determined whether the number of the maximum reconstruction trees is unique.When the largest reconstructed tree is unique,the tree is the optimal reconstructed tree;if the largest reconstructed tree is not unique,the difference in computing power between different reconstructed trees is judged,and the combination with the strongest computing power is selected for reconstruction.Finally,the bottleneck bandwidth of different reconstruction methods is compared through experiments,and the reconstruction time of different participating nodes is compared to illustrate that the bottleneck bandwidth of the TRBCC reconstruction method is large,the reconstruction time is short,and the performance is superior.