Research on Block Device Driver for I/O Scheduling of Solid-State Storage Array

The I/O performance of storage devices has a decisive influence on the I/O performance of the whole parallel storage system. Because the file system is usually installed on the block device driver, the access to the block device driver is very frequent, so the performance requirements of the block device driver are very high. In this paper, the block device driver of solid-state storage array I/O scheduling is studied, and an adaptive partition-based I/O scheduling mechanism for solid-state storage array is proposed and implemented. This scheduling mechanism will partition all I/O requests in the current time period according to the starting block number before I/O processing, and I/O requests belonging to the same partition are most likely to be associated with a specific application. Then, after the partition is completed, I/O requests in each interval are merged and reordered. The results show that for Tracel and Trace3, the performance of Adaptive partition is improved by 28.52% and 59% compared with Non-merging. The experimental results show that the new I/O scheduling algorithm based on adaptive partition can reduce the overall I/O completion time, especially the time loss caused by merging continuous I/O requests.