一种基于分布式存储系统的Piggyback码

Piggyback Code Based on Distributed Storage Systems

随着大数据时代的来临,纠删码在分布式存储系统中有着越来越重要的应用.Piggyback码作为纠删码的一种,因其同时具有高效率存储和低修复带宽的优点而成为近年来的研究热点,RSR-Ⅱ码作为Piggyback码中在减少修复带宽方面最典型的码,因其修复过程中需要进行有限域上方程组的求解,使得编码复杂度和修复复杂度过高.针对这个问题,提出了一种新的Piggyback码,并给出了其一般性构造和修复算法,该码基于分布式存储系统中广泛使用的系统型M DS码,通过结合Piggybacking框架的核心思想,构造了新的piggybacks添加规则,有效避免了有限域上的方程组求解问题.对比分析表明,新的Piggyback码既保持了RSR-Ⅱ码较低的平均修复带宽率,还具有更低的编码复杂度和修复复杂度.

With the advent of the big data era,erasure codes becoming more and more important in distributed storage systems.As a kind of erasure codes,the Piggyback codes have high efficiency storage and low repair bandwidth have become a research hotspot in recent years.The RSR-Ⅱ code as the most typical code in the Piggyback codes for reducing the repair bandwidth,but it needs to solve the equations in the finite field during the repair process,which makes the encoding complexity and repair complexity high.To solve this problem,we propose a new Piggyback code and give its general construction and repair algorithm.The new Piggyback code bases on the systematic MDS code which widely used in distributed storage systems,and combines with the core idea of Piggybacking framework,we construct a new rule to add piggybacks.After the new design,the problem of solving the equations in the finite field is effectively avoided.The comparison analysis shows that the new Piggyback code not only maintains the lower average repair bandwidth rate of the RSR-Ⅱ code when repairing the failed systematic nodes,but also has lower encoding complexity and repair complexity.