Construction Of Fractional Repetition Codes In Distributed Storage Systems

In the context of the era of big data,how to effectively store massive amounts of data has become a current research hotspot.Traditional centralized storage has temporarily solved this problem,but its storage cost is high and there are system performance bottlenecks which making distributed storage systems gradually replace it as the first choice for massive data storage.Distributed storage systems have the characteristics of high scalability and low cost.At present,the research on distributed storage systems mainly focuses on how to repair storage nodes when they are failed,and ensure the reliability of distributed storage systems.In distributed storage systems,common failure node repair strategies include replication strategy and erasure coding strategy,but both of them have their own limitations.The replication strategy requires too much storage overhead,and the erasure coding strategy requires too much bandwidth to repair the failed nodes during the repair process.The proposal of Fractional Repetition(FR)codes avoids the above limitations to a certain extent.FR codes have lower repair locality and repair bandwidth costs in the process of repairing failed nodes,and can perform no coding quick fixes on failed nodes.These characteristics make FR codes get a lot of attention since it was proposed.To this end,this paper has conducted research on the construction algorithm of FR codes in distributed storage systems.The main contents are as follows:(1)Aiming at the effective repair and different storage capacity of fractional repetition codes,a new coding scheme of fractional repetition codes based on Harary graph is proposed.First,fill the edges of the Harary graph with data blocks according to the law,and transform it into an incidence matrix to obtain a fractional repetition code.and further we use spanning trees and eccentricity to construct a new code called FRSH(Fractional repetition based on spanning trees of Harary graph)codes.FRSH codes can achieve heterogeneous storage capacity of nodes,compared with the existing RS(Reed-Solomon)codes and Simple Regeneration Codes(SRC),FRSH codes has better performance in repairing bandwidth overhead and repairing locality,moreover,it also improve the repair efficiency and repair time of the failed nodes.(2)In order to improve the repair efficiency of multi-node failure in distributed storage systems,a new FR code called Fractional Repetition based on Regular Matrix(FRRM)code is proposed.The code construction algorithm further realizes the heterogeneity of data block repetition on the basis of the capacity heterogeneous FRSH code.Specifically,firstly,an FR code with homogeneous node capacity is constructed based on a single regular matrix,and the FR code obtained by this construction algorithm has universally good characteristics.In order to meet the different storage requirements of hot and cold data in the actual storage environment,the concatenated matrix of FR codes with different repetitions is cascaded diagonally through matrix,which can extend to FR codes with heterogeneous data block repetition and node capacity.In this heterogeneous FR code,hot data with higher access frequency has higher repetition,and cold data with lower access frequency has lower repetition.According to the change of access frequency of cold and hot data,the repetition of cold and hot data in FR code can be adjusted by adjusting matrix.Compared with RS codes,SRC and traditional FR codes,the constructed FR code has lower overhead in terms of computational complexity and repair bandwidth overhead,and has better storage efficiency and system adaptability than the traditional FR code.