Research On The Distributed Storage Systems Based On Network Coding

The concept of network coding is proposed by R.Alswede in Network Information Flow for the first time, which has drastically challenged the conventional store-and-forward routing mode, for the reason that it allows intermediate nodes to restructure the packages received, and therefore makes it to reach the upper bound capacity of the multicast communication network. Acedanski has surprisingly introduced the idea of network coding into storage technology, more specifically saying, random linear coding scheme, in How Good is Random Linear Coding Based Distributed Networked Storage. Since it has been put forward, it has attracted high attention among scholars.In recent years, with the rapid development of Internet, network information has been thriving in exponential growth, and thus information technology industry has turned its focus to the study of storage technologies. The traditional Client/Server mode can no longer satisfy people’s increasing needs. Nonetheless, distributed storage system substantially makes use of the idle network storage space, and assembles the idle Internet PCs together to build an information storage resource pool. It has gained extensive application for the sake of its good extensibility and low price. Moreover, network coding is even suggested for distributed storage system. Researches show that distributed storage system with network coding can not only save bandwidths, but also balance the load well, as well as enhance the system robustness.This paper first presents the study on storage schema based on random linear coding, in comparison with common storage schemes, i.e. copy and erasure codes. Conclusion can be drawn that storage schema based on random linear coding consumes less redundant storage bandwidth while restructures source file in high probability. Then, specific to distributed storage system based on network coding schema, this paper conducts investigation of node repair technology on regenerating codes. The best advantage of regeneration code is that failure nodes can be repaired in no need of downloading the whole source file. So it can reduce the use of the repair bandwidth. Relevant literature proves that there exist two extreme points in regeneration method: Minimum Storage Regeneration (Minimum storage regeneration points, MSR) and Minimum Bandwidth Regeneration (Minimum bandwidth regeneration points, MBR). Correspondingly, here in this paper, research on two node-repairing technologies is exerted:Minimum Storage bandwidth Regenerating Codes (Minimum storage regeneration code, MSRC) and Minimum node Bandwidth Regenerating Codes (Minimum bandwidth regeneration code, MBRC). Calculation results show that the node-repairing technology based on MBRC can minimize the consumption of bandwidth when repairing the failed nodes. At the end of this paper, a detailed analysis of the MBRC schema has been made, and simulate the data distribution, data reconstruction and the repair of the failure nodes of MBRC, on the platform of VC++6.0. It’s verified that the MBRC schema is effective, as a network coding schema.