An efficient I/O scheduler over Multi-Bank Flash Memory storage systems

Abstract

Flash memory has been widely used in embedded systems because of its advantages. The traditional scheduler over flash memory storage systems, namely NOOP, makes no distinction between read and write requests. There is much room for improving the I/O performance, especially over multi-bank flash memory storage systems. In this paper, a novel scheduler, namely Multi-Bank Flash Memory scheduler (MBF scheduler), is proposed to take advantage of the parallelism of multiple banks. MBF scheduler reorders read and write requests and assigns higher priority to read requests. The experimental results show that I/O throughput and average response time are improved significantly compared with NOOP, and MBF provides rather even erase counts and capacity utilizations among different banks.