EA-EO: Endurance Aware Erasure Code for SSD-Based Storage Systems

Abstract

One of the main issues in Solid State Drive (SSD)-based storage systems is endurance which is directly affected by the number of Program/Erase (P/E) cycles. The increment of P/E cycles increases the bit error rate threatening the reliability of SSDs. Erasure codes are used to leverage the reliability of storage systems but they also affect the number of P/E cycles based on their code pattern. A lower dependency between data and parities in the code pattern may lead to smaller number of P/E cycles providing better endurance. This paper introduces an Endurance Aware EVENODD (EA-EO), which minimizes the dependency between data and parities in the coding pattern. A simulation environment is used to compare the write-cycles of EA-EO with EVENODD in terms of different request size. The results show that the endurance improvement of EA-EO code could be as high as 44%. Furthermore, performance analysis of these codes in terms of parity construction and failure recovery shows that the number of XOR-operations is reduced in EA-EO compared to EVENODD.