Study On Redundancy Strategies Based On Network Coding In Distributed Storage Systems

Data have exploded in recent decades,distributed storage systems(DSSs)has become the preferred platform for storing massive data due to its low cost and high performance.However,in the actual distributed storage systems,node failures occur frequently,resulting in the loss of useful information.In order to ensure the availability and reliability of useful data,it is important to introduce data redundancy into the distributed storage systems.Replication is the simplest redundancy method,data replication supports efficient repair and is easy to manage in practice,yet it suffers from the drawback of high storage overhead.Instead,the erasure codes strategy greatly improves the storage efficiency,but requires a large amount of network resources during node recovery.For this reason,Dimakis et al.applied the network coding technology to distributed storage,minimizing repair traffic by allowing intermediate node coding,thereby reducing the bandwidth overhead of failed node repair.Based on network coding theory,regenerating codes attempts to minimize the amount of data download required for node repair and thus reduce the repair bandwidth overhead,while locally repairable codes(LRCs)aims to minimize the number of nodes accessed during node repair and thus reduce the repair locality.Fractional repetition(FR)codes is a special minimum-bandwidth regenerating(MBR)codes,which greatly reduces the complexity of node repair due to its exact non-coding repair characteristics.Based on the network coding theory,this article proposes three coding methods,the main contents are as follows:(1)The existing LRCs with(r,t)-locality are rarely able to achieve optimal minimum distance and optimal code rate at the same time.In this article,an optimal LRCs with availability t =2 is proposed based on Latin square.In order to achieve arbitrary availability,an optimal LRCs based on orthogonal Latin squares is further proposed.Theoretical analysis shows that both LRCs proposed in this article achieve optimal minimum distance,and the LRCs based on Latin square achieves optimal code rate.(2)Considering that the construction parameters selection of current LRCs is not flexible enough to adapt to the dynamic parameter changes in distributed storage systems,the LRCs with arbitrary(r,t)-locality is further constructed.In this article,we first construct an all symbol-locally repairable code with arbitrary(r,t)-locality based on iterative matrix,and then improve the structure of its check matrix to construct an information symbol-locally repairable code with arbitrary(r,t)-locality.Theoretical analysis shows that both two LRCs proposed in this article can not only achieve arbitrary locality and arbitrary availability,but also have optimal code rate and optimal code length when t =2.(3)In order to further reduce the complexity during node recovery,a construction algorithm of FR codes by cyclic permutation matrix is proposed in this article.The construction parameters of current FR codes are limited,the coding scheme proposed in this article can not only construct FR codes with arbitrary data repetition degree and node storage capacity,but also realize real-time repetition degree adjustment of cold data and hot data through matrix transformation.Theoretical analysis shows that the constructed FR codes with good repair locality and repair selectivity is generally FR codes,under certain parameters,it also achieves the optimal minimum distance and optimal reconstruction degree.