On fixed-priority schedulability analysis of sporadic tasks with self-suspension

We consider the schedulability analysis problem of a set of sporadic tasks which are subject to self-suspension, using a fixed-priority scheduler on a preemptive uniprocessor. We show that this problem is coNP-hard in the strong sense even in the simple case when only the lowest-priority task is self-suspending. Also, it is shown that the problem is weakly coNP-hard even if that self-suspending task has only a single suspension interval. In addition, we propose an efflcient me- thod for schedulability analysis of self-suspending tasks that are subject to interference from higher-priority tasks without self-suspension. The method works on the basis of an itera- tive approach which begins with an abstraction of the task set and improves the analysis results by refinement steps as needed. Our evaluation shows that this method significantly improves the scalability of the existing approaches.