Can Erasure Codes Damage Reliability in SSD-Based Storage Systems?

Abstract

Erasure codes are applied in storage systems including both Hard Disk Drive (HDD) and Solid State Disk (SSD) to protect arrays of disks against failures. Applying these codes in SSD-based systems incurs additional number of Program/Erase (P/E) cycles on each disk, which may accelerate the wear-out of disks. This means that while erasure codes improve reliability of SSD-based systems, they impose a side-effect that may degrade reliability as the number of P/E cycles increases. This paper investigates the benefit and side-effect of erasure codes on reliability of SSD-based systems. The investigation attempts to find out the parameters which improve/damage reliability. This study has been evaluated using the DiskSim and SOYA simulators, and exploiting different configurations for a storage system running nine real workloads. In the simulation, the effect of five parameters on the improvement/damage of reliability is studied; these parameters are mapping P/E cycles with failure rate, disk failure rate, disk repair time, number of disks, and coding pattern of erasure codes. Simulation results show that in almost all cases, the effect of erasure codes on reliability improvement is on average 16 percent greater than their effect on reliability degradation.