DYNAMIC SCHEDULING OF REAL-TIME TASKS UNDER PRECEDENCE CONSTRAINTS

While scheduling theory has been developed over a long period of time, it is important to note that most results concern problems with static characteristics. However, a real-time system is dynamic and requires on-line and adaptive scheduling strategies. So, an important aspect of real-time systems research is to devise methods flexible enough to react to a dynamic change of processor load and to attempt to schedule all the tasks judiciously. In this paper, we are particularly concerned with the problem of scheduling tasks which are of two kinds: periodic and sporadic, on a monoprocessor machine. Periodic tasks are independent, run cyclically and their characteristics are known in advance. In addition, we allow for the unpredictable occurrence of aperiodic task groups, with timing and precedence constraints. Clearly, the main problem is to devise a schedulability test that makes it possible to decide whether a new occurring group can be accepted, without upsetting the tight timing behavior requirements. We present an optimal acceptance test which returns a decision on the basis of the current state of processor load and by considering tasks to be scheduled according to the well known preemptive algorithm Earliest Deadline. The acceptance algorithm and a complexity analysis are presented. © 1990 Kluwer Academic Publishers.