A Flexible and Low-Complexity Local Erasure Recovery Scheme

For large-scale distributed storage, besides redundancy, locality in terms of the number of data and parity symbols to access for failure recovery is a critical issue to address in order to ensure data availability, lower network traffic, and reduce recovery latency. A flexible yet low-complexity scheme for local erasure recovery is presented in this letter. Although the proposed scheme is not maximum distance separable, it does not have any constraint on the locality nor the parameters of the code, allows easy tradeoff on the locality and redundancy, and can achieve unequal protection over drives with different reliabilities. More importantly, it enables low-complexity implementation. An efficient hardware architecture for the proposed scheme is developed. It has very small overhead compared with traditional erasure codes when correction capability is not large, such as 2, 3, or 4.