The Failure Repair Approach For Distributed Storage System Based On Network Coding

Nowadays, human society has been enterring a period of rapid development of information technology, and information plays a vital role in the social life. Meanwhile, massive data storage is also growing attention. Especially in recent years, the wide application of high-definition videoes and big data, the traditional centralized data storage systems have been showing a number of disadvantages, such as low security, low reliability, weak scalability etc. Fortunately, the distributed data storage systems can make up for those shortcomings. They allow massive amounts of data to be stored in the distribution form to meet the massive capacity needs, but also have a higher stability level.However, distributed data storage systems have to be faced with the problem of the storage nodes failures. So, it has become an important issue in distributed data storage systems that how to repair the failure nodes to guarantee systems running in a stable way. Currently, the key indicators of repairing failure nodes include nodes storage capacity, repair bandwidth, repair cost, repair time, etc. It has become hot research topic that how to repair failure nodes while reducing the related indicators. This paper studies the distributed data storage based on network coding and failure repair strategies, the main work and contributions are as follows:(1) In this paper, the failure repair problems are abstracted into network coding based on the information flow diagram, which will simplify the analysis of data transmission process with the related graph theory.(2) The theoretical analysis of the lower bound of the repair time is made, considerring the current repair approaches are not optimal in terms of rebulid time. In this paper, we propose MTR approach (Minimum-Time Repair), which can achieve the lower bound of the repair time. And, on this basis, the repair bandwidth is reduced to a minimum value. The numerial anaysis shows that MTR repair strategy can greatly reduce the repair time in comparison to current repair approaches. It is very important for distributed data storage systems that need to be repaired rapidly. (3) Taking the constraint repair time into account in the practical applications, we propose the optimal repair-bandwidth repair strategy in the constraints of repair time. We also bulid the optimization model of the strategy and give the optimal solutions explicitly. This approach can repair failure nodes within the given threshold time, and repair bandwidth on the corresponding point is minimal. Meanwhile, the strategy is flexibile, in any repair threshold time (greater than the lower bound of repair time), there exists repair strategies which optimize the repair bandwidth.