Using SATF Scheduling in Real-Time Systems
Lars Reuther, Martin Pohlack · 2003
Introduction Disk scheduling algorithms based on the rotational position of the disk head (Shortest Access Time First scheduler, SATF) are known to be a good approximation of an optimal disk scheduling algorithm [3]. However, because of the required knowledge of the disk they are also believed to be difficult to implement outside of the disks firmware at driver level [2]. Unfortunately, scheduling cannot always be done in the disk drive. With queueing a large number of requests in the disk drive, the driver resp. the operating system loses the ability to control the point in time a single request is executed. But this control is essential for certain systems, e.g. if a system must meet deadlines of disk requests. Another drawback is the limited number of requests a disk can queue, we experienced 32 to 64 requests with current disks. However, scheduling algorithms work better the larger the number of requests they can choose from. The aim of our work is to (a) show the feasibility