Evaluation of Scheduling Algorithms for Real-Time Disk I/O

Real-time disk I/O scheduling is extremely important to the performance improvement of the whole real-time system since the disk devices are the system’s bottleneck. To meet time constraints, some conventional algorithm, such as EDF, Least Slack Time First, Rate Monotonic, can be used to schedule disk I/O. However, these algorithms ignore the relative position of requested data on the disks, hence their throughput is certainly low due to large arm positioning time. To keep a good tradeoff between optimizing throughput and meeting time constraints, several hybrid real-time scheduling algorithms were proposed. These algorithms include Priority SCAN(P-SCAN), SCAN-EDF, Feasible Deadline SCAN(FD-SCAN), Feasible Deadline EDF(FD-EDF), and Shortest Seek Earliest Deadline by Order/Value(SSEDO, SSEDV). In this project, we compared real-time algorithms, such as EDF, FD-EDF, SSEDO and SSEDV, with conventional algorithms, such as FCFS, SSTF and SCAN, in the aspects of guarantee ratio, average throughput and merit according to our simulation results.