Low-Complexity Shorten MSR Code for Limited Bandwidth Systems

Ke Li,Shushi Gu,Ye Wang,Qinyu Zhang
DOI: https://doi.org/10.1109/ICCChina.2018.8641182
2018-01-01
Abstract:Regenerating code is a valid approach to improve the reliability and availability in distributed storage system (DSS), considered as the optimal tradeoff between repair bandwidth and storage amount. Nevertheless, in some realistic practical network system, due to the limited nodes and bandwidth resources, redundant bandwidth consumption and assisted nodes connections will be essential indeed. Aim to the problem of regenerating code applying to limited resource systems, we propose a novel distributed coding scheme, termed shorten minimum storage regenerating (sMSR) code with two critical targets, unit storage cost and unit repair bandwidth. The construction of sMSR is also given by removing some information bits whose mother code is generated by product matrix in encoding process. This code could cut down the number of nodes occupied and minimize unit bandwidth cost for repairing failed nodes. Additionally, we implement binary addition and shift implementable convolution (BASIC) on our code to decrease the computational complexity. It demonstrates that our coding method outperforms existing conventional RGCs with fewer bandwidth overhead and higher repair efficiency.
What problem does this paper attempt to address?