Abstract

We consider the "supertask" approach, proposed by Moir and Ramamurthy at RTSS '99 as means for supporting non-migratory tasks in Pfair-scheduled systems. In this approach, tasks bound to the same processor are combined into a single supertask, which is scheduled as an ordinary Pfair task; when a supertask is scheduled, one of its component tasks is selected for execution. Unfortunately, while Moir and Ramamurthy's paper suggests that supertasking is a promising approach, counterexamples presented by them show that non-migratory tasks can actually miss their deadlines when supertasking is used in conjunction with all known Pfair scheduling algorithms. In this paper, we show that such deadline misses can be prevented by inflating each supertask's utilization. We present experimental evidence that shows that the required inflation factors should be small in practice. We also show that these ination factors usually can be reduced if, instead of being scheduled in a Pfair manner, non-migratory tasks are scheduled (perhaps "unfairly") using earliest-deadline-first priorities. In soft-real-time and rate-based systems, it may be permissible for a task to miss a deadline by a small amount. We show that our ination factors also can be reduced if this is the case.

Citations