Research On Repair Node Selection In Distributed Storage System Based On Network Coding

With the rapid development of information technology, global data storage presents explosive growth, therefore how to efficiently store massive data becomes an urgent problem to be solved. Due to the advantages of quick, high scalability and so on, distributed storage system has been widely applied to massive data storage. In order to improve the storage reliability, the system stores redundantly data and provides a positive node repair mechanism. Because the consumption of node storage and repair bandwidth is very large by using traditional redundancy strategies(replication, erasure codes), network coding is introduced into the distributed storage system. Although network coding can bring lots of benefits, node repair will consume a lot of unnecessary regeneration time.In this paper, we study the problem of repair node(new comer, provider) selection in distributed storage system based on network coding. Firstly, the paper studies the provider selection problem with the optimization goal of reducing regeneration time(The new comer is given). When the scale of problem is small(The number of storage servers is little), the paper proposes a mixed integer linear programming with constraining the provider selection problem with linear programming; when the scale of problem becomes big and the number of nodes becomes more, the paper relaxes the integer linear constraint and proposes a heuristic selection algorithm to select approximately optimal providers and data transmission paths from the providers to new comer, which is suitable for small data centers. Secondly, to further reduce regeneration time, we consider the joint selection problem which is equivalent to the selection problem of the new comer and provider, then we propose a kind of heuristic selection algorithm by the methods of modeling the selection problem by relaxing repeatedly integer constraints to select approximately optimal providers、new comer and data transmission paths, which is suitable for large data centers. Finally, the above two schemes only consider how to reduce regeneration time, in the actual network, we should consider data transmission cost of node repair under keeping minimum regeneration time in repair process, therefore we propose transmission cost optimization algorithm with constraining transmission paths from providers to new comer.The simulation results show that the proposed algorithms can effectively reduce regeneration time and improve the transmission efficiency. In addition, the transmission cost optimization algorithm designed in this paper can effectively reduce the transmission cost. Consequently, this research for the further development of distributed storage system has positive significance.