Research On Exact Repair Of Regenerating Code Based On Distributed Storage System

With the continuous progress of science and technology,the development of information technology has been out of our imagination.Frequent network activities lead to the rapid increase of data generated by network systems.Due to massive storage data,traditional centralized network storage has been difficult to meet the growing demand for large-scale storage space.Distributed storage system is widely used and developed because of its massive storage capacity,high scalability and low cost advantages.With the development and widespread application of distributed storage systems,the problem of data reliability caused by this hidden danger becomes more and more serious.Therefore,how to ensure the reliability of data has become a"top priority" in the development of distributed storage system.Ensuring the reliability of data mainly depends on data fault-tolerant technology,and the key to solve data fault-tolerant technology is effective data repair.That is,when some storage nodes fail,the surviving nodes consume as little system resources as possible to complete the repair of invalid nodes.Regenerating codes based on network coding have been widely studied for their less redundant storage,high reliability and low repair bandwidth.However,there are some shortcomings of the current study in distributed storage regeneration codes.Various regenerating codes based on distributed storage systems have different tolerance degree to storage nodes due to their own structure and other reasons.The smaller the tolerance degree is,the lower the data reliability and security of storage systems,and the low tolerance degree of some regenerating codes leads to fewer repairable nodes.Although other existing regenerating codes have high tolerance degree,they need a storage system with high storage capacity,high repair bandwidth or high computational complexity,so they bring about too high cost.We need to design a storage system with high tolerance degree,lower repair bandwidth,lower storage capacity and lower computational complexity.In order to solve the above problems.based on fractional repetition codes,this paper proposes a Multi-repair Fractional Repetition Regenerating Code(MFRRC)and a Heterogeneous Fractional Repetition Regenerating Code(HFRRC).The main work and innovations of this paper are summarized as follows:In order to solve the problem that only one storage node can be allowed to fail in previous fractional repetition codes,a Multi-repair Fractional Repetition RegeneratingCode is proposed in this paper.This research mainly includes the construction of the code,the analysis of the process of accurate data repair,the establishment of the priority of node selection in the repair process and the performance analysis.We introduce the concepts of parent storage nodes and descendant storage nodes in this coding style.In other words,we add corresponding storage copies of the original storage nodes in order to ensure multiple failed storage nodes can be all exact repaired.Through theoretical analysis,we conclude that this encoding:,method inot only improves the tolerance,but also maintains the advantages of the original fractional repetition regenerating codes.At the same time,we make simulation and numerical analysis of this encoding method by setting relevant parameters.The simulation results show that the MFRRC we proposed can greatly improve the tolerance dergree,and can reduce the total repair bandwidth and total repair time effectively.Aiming at the problem that the number of help nodes is larger and only one storage node is allowed to fail in the previous fractional repetition regenerating codes,we proposed a Heterogeneous Fractional Repetition Regenerating Code in the case of heterogeneous distributed storage systems(i.e.,the amount of data downloaded from the help nodes surviving is different in the process of repair).The contents of this research mainly include the code construction,the performance analysis and the existing problems.To optimize this coding method,we propose an extended structure.Firstly,in this coding method,we adopt the coding structure of heterogeneous repair bandwidth,which can reduce the number of helper nodes involved in the process of exact data repair.Secondly,in the extended structure of HFRRC.the concepts of parent storage node and child storage node are introduced again,which greatly improves the tolerance.In addition,we simulate and analyze to compare the performance in tolerance degree or other performances of this encoding method and other typical encoding methods.The results of simulation show that the HFRRC we proposed not only Greatly reduces the number of help nodes and improves the tolerance decree,but also has a small total repair bandwidth and a fast and efficient data repair process.