Study On Fast Repair Of Failed Nodes In Distributed Storage Systems

With the advent of information age and the development of Internet technology,the global data is growing explosively.Distributed storage systems have been widely used by individuals and companies due to its high scalability and low cost,gradually replacing the traditional centralized storage systems.However,storage devices in distributed storage systems are inexpensive,unstable and will inevitably fail,which requires frequent and quick data repair operations.Thus,it is necessary to ensure the reliability and availability of data by storing redundant data.At present,the most common redundant mechanisms are replication and erasure codes.Replication needs to store copies of the original file,resulting in larger storage overhead.Erasure codes have better storage overhead while bringing higher repair bandwidth overhead.Therefore,in different storage environments,it is an important issue that we adopt the best encoding methods to repair the failed nodes quickly.The main contents of this paper are as follows:(1)In order to improve the reliability and repair efficiency of the distributed storage systems,this paper proposes a class of Minimum Bandwidth Regenerating(MBR)codes based on Variable Fractional Repetition(VFR)codes,which can repair the failed node quickly and accurately.Specifically,considering the imbalanced of data accessed by users,the cyclic VFR codes are constructed according to the idea that data with different heat degrees are copied in different repetition degrees.On this basis,the cyclic VFR codes are grouped to improve the file download speed,and the corresponding MBR codes are achieved.When the MBR codes based on cyclic VFR codes are adopted in distributed storage systems,the repair locality of a single node failure is also 2,which improves the defect of traditional MBR codes with much larger locality.Compared with the existing RS codes and simple regenerating codes,the proposed MBR codes have lower repair locality,repair complexity and repair bandwidth overhead,moreover having higher repair efficiency.Meanwhile,the codes expand the application scopes of traditional FR codes to a certain extent.(2)To ensure load balancing of nodes,a class of Rotating Interleaving Hierarchical Codes(RIHC)is proposed to repair the failed nodes quickly.Firstly,the Rotating Interleaving Codes(RIC)are constructed.By using the parallel I/O property of the nodes during repairing a single failed node,the amount of data transmitted by each help node is the same,that is,the node load is balanced.Then taking the RIC as the basic group,and the RIHC are obtained by rotating and expanding according to the construction method of the Hierarchical Codes(HC).Performance analysis and simulation experiments show that,compared with the existing HC and RS codes,RIHC have lower repair locality and repair bandwidth overhead,and only adopt simple XOR operations to reduce the repair complexity.