Study On Distributed Storage Redundancy Strategies For The Repair Of Failure Nodes

With the rapid development of mobile technology and the universal application of various mobile devices,a large amount of data is generated every day.How to store and extract these massive data is one of the challenges we face today.Distributed Storage Systems have gradually become the preferred storage systems for big data storage today,with its own advantages such as low investment cost,strong manageability,and good shareability.When a distributed storage system faces disk data loss,redundancy strategies need to be introduced to ensure that the system can work normally.The locally repairable codes(LRC)use the generated local check block to repair the failed node,which greatly reduce the repair locality.Fractional Repetition(FR)codes are a type of minimum-bandwidth regenerating codes,which can repair faulty nodes accurately and without coding.In order to ensure that the storage systems can still work normally in the case of multiple node failures,this article proposes three coding constructions,the main contents are as follows:(1)Considering the large amount of calculation for repairing regenerating codes,LRC is introduced,which reduce the amount of calculations and repair locality in node failure repair.In the existing graph-based LRC,the parameter limit is relatively large.This paper proposes a hypercube-based LRC with availability t(28)2,which mainly use the distance function between vertices to associate a point with a distance of 1 and a point with a distance of 2,and combines the identity matrix to construct the check matrix.It also proves that the proposed LRC have the optimal minimum distance and optimal code rate.(2)Since the LRC with availability t(28)2 can only tolerate two node failures,an LRC with arbitrary(r,t)locality is further constructed.First,method one uses non-cyclic relative difference sets to construct a incidence matrix that meets the conditions.Method two uses a Unital design to construct a incidence matrix.Finally,the incidence matrix constructed above and the identity matrix are combined to generate a check matrix,so that each information bit is local.Each repair group of the LRC constructed by these two construction methods contains only a single parity check symbol,and the minimum distance is the optimal.(3)In order to effectively reduce the repair complexity and repair bandwidth overhead during node fault repair,this paper presents a coding scheme using difference matrix to construct FR codes.The FR codes coding scheme uses a difference matrix and Kronecker sum to construct an orthogonal arrangement.According to the orthogonal arrangement,each row of the same element is taken as the coding block of the node to obtain the corresponding FR codes.The constructed FR codes can be completely decomposed,that is,it can be divided into multiple parallel classes,and at the same time,the repeatability of each data block and the storage capacity of the node can be adjusted.In order to reduce the repair locality,you can construct an FR codes with less locality by adjusting the parameter ? of the difference matrix.