Research On Data Reconstruction Optimization Based On LRC

With the advent of the era of big data,erasure codes as a general fault-tolerant mechanism for distributed storage systems are also facing new challenges.The advantage of erasure codes is that storage utilization is very high,but a large amount of data transmission is required to recover data.However,these studies tend to focus on single-node data repair optimization without discussing the failure of multi-node data.The reason is that researchers generally believe that the probability of single-node failure is much more than the failure of multi-node data.However,as the size of storage centers continues to increase,the probability of simultaneous failure of multiple nodes increases,and multi-node data reconstruction has already occurred.Become a major factor affecting the overall performance of distributed storage systems.In order to provide better data repair performance,especially multi-node data repair performance,a Rotated Group Repairable Codes(RGRC)is proposed based on the LRC,which proposes a full packet and overlap coding strategy,which can reduce data repair cost.Under the same fault tolerance,RGRC reduces cost of single node repair by 50%compared to RS code,and the multi-node repair cost is reduced by 25%.Compared with the LRC,the single-node repair performance of the RGRC code is slightly better than the LRC.Multi-node repair costs are reduced by 25%.In order to make actual comparison and application of RGRC,a(12,3,2,1)RGRC is deployed in Jerasure,It was then tested with a(12,3,2)LRC and a(12,4)RS code with the same fault tolerance.The test results show that the(12,3,2,1)RGRC code has improved by 50% in data repair performance over the(12,4)RS code in the event of a single node failure,compared to the(12,3,2)LRC code.The data repair performance has improved by about 15%.When multi-node data fails,the(12,3,2,1)RGRC improves performance by 60%-70% compared to the(12,4)RS code,while the contrast(12,3,2)LRC,Repair performance has also improved by about 30%.After that,a parameter comparison experiment was also carried out.In the case where the redundancy is 6,the(12,2,3,1)RGRC is compared with the(12,3,2,1)RGRC and(12,2,2,2)Repair performance of RGRC.As a result,we found that more local groups can betteruse the intra-group coding method for data reconstruction,thus reducing the cost of data repair.