Abstract
Although the advances in hardware technology have dramatically increased processor speeds recently, the access time of storage devices does not decrease rapidly when disk capacity is improved at 60－80% annually. As a result, the gap between the performance of processors and the performance of disks is enlarged significantly and the bottleneck of the performance in a computer system is shifted to the storage subsystem. Real-time disk scheduling is an important research topic for time-critical multimedia applications. For real-time disk scheduling, our goal is to minimize disk access time while guaranteeing tasks’ timing constraints. Some well-known research results, such as SCAN-EDF (SCAN-Earliest-Deadline-First) and DM-SCAN (Deadline-Modification-SCAN), in the DM-SCAN approach, a deadline modification scheme was applied, and the modified deadlines were smaller than the original one, applied the SCAN scheme to reschedule service sequence of input tasks to reduce tasks' service time. The performance is limited as the seek-optimizing SCAN scheme is only applied to tasks with the same deadline or within a confined group. For example, in DM-SCAN, groups of tasks that can be successfully rescheduled by SCAN under specified real-time requirements are selected. In addition, input schedule must be in Earliest-Deadline-First (EDF) sequence. In DM-SCAN, a deadline modification scheme was employed to obtain a pseudo EDF sequence from non-EDF ordered input tasks. Because the modified deadlines were earlier than the original ones, the number of tasks that could be rescheduled is decreased and thus data throughput is reduced. In this paper, to resolve the drawback of DM-SCAN, we propose a new real-time disk-scheduling algorithm that uses the concept of Shortest-Task-First with deadline modification. We applied the DM scheme carefully to improve the overall performance. As shown in experimental results, our approach indeed has good performance and makes more tasks to meet their deadlines. In addition, we also give a formal proof to show that the general Disk-scheduling problem with linear cost-function is NP-Hard.