Research Of Regeneration Codes In Distributed Storage Systems

With the development of cloud computing,data storages have been the key techniques in cloud storage systems.It is essential to ensure the reliability and validity of the data,as well as the good tradeoff between the storage space and bandwidth consumption in the cloud storage systems.Distributed storage systems usually adopt to add redundancy to guarantee integrity and reliability of data.These data are divided into multiple blocks which are stored in massive storage nodes of the networks.The earliest form of redundancy is repetition strategy.The multiple copies of the data would be stored in different nodes.When one node fails,the storage system can fetch the failure data from the other available nodes that store the redundant replicas.However,repetitive mechanism would cause large storage consumption because of the simple repetitions.Obviously,the low efficient repetition can no longer meet the explosive increasing demand of big data storage.The erasure code is a popular solution to improve storage efficiency in the face of network node failure.Erasure codes require less redundancy than simple replication for the same level of reliability.Although erasure code provides a large storage benefits,the bandwidth cost is too high to be practical for distributed storage system.To reduce repair bandwidth,regenerating codes are subsequently presented.Regeneration codes can achieve the optimal tradeoff curve of storage space and repair bandwidth in a single node repairing,but it is suboptimal when multiple nodes are damaged at the same time.Because regenerating code can only individually repair a single node each time with one-by-one manner,it should result in the higher total repair bandwidth of regenerating code.Recently,cooperative regenerating codes have been paid more attention because they can simultaneously repair multiple nodes and they have the better tradeoff curve of storage space and repair bandwidth than regeneration codes in the repair of multiple nodes.But for cooperative regenerating codes,their constructions are inflexible and their parameters are based on particular value.The main works of this paper are as follow:(1)We have researched replication strategy and coding strategy such as erasure codes,regeneration codes and cooperative regenerative codes.(2)We analyzed the advantages and disadvantages of the existing two MSR codes which based on product-matrix.We also designed and deduced a low complexity of MSR codes based on product-matrix to reduce the complexity of encoding and decoding proceeding.We illustrated the procedure of the encoding,the failed node repairing and the data reconstruction.(3)We analyzed the advantages and disadvantages of the existing MSCR codes,and then deduced a kind of novel MSCR codes constructions from PM_MSR codes to improve the shortcomings of the less flexible parameters.(4)Base on JAVA programming,we have implemented the two proposed algorithms in this paper,and compare with the existing algorithms.The experimental results show that the proposed schemes are feasible and effective.