On the parallelism of I/O scheduling algorithms in MEMS-based large storage systems

MEMS-based storage is being developed as a new storage media that has several salient characteristics such as high-parallelism, high density, and low-power consumption. Because physical structures of MEMS-based storage is different from those of hard disks, new software management techniques for MEMS-based storage are needed. Specifically, MEMS-based storage has thousands of parallel-activating heads, which requires parallelism-aware request scheduling algorithms to maximize the performance of the storage media. In this paper, we compare various versions of I/O scheduling algorithms that exploit high-parallelism of MEMS-based storage devices. Trace-driven simulations show that parallelism-aware algorithms can be effectively used for high capacity mass storage servers because they perform better than other algorithms in terms of the average response time when the workload intensity becomes heavy.