Resource Bounding for Non-Preemptive Task Scheduling on a Multiprocessor Platform

Abstract

Task scheduling, which is the fundamental problem for real-time systems, has been approached from various points of view and for various classes of hardware/software configurations. Most of the results currently available have been determined for preemptive scheduling. However, the non-preemptive case is also of great interest, and its higher complexity requires different solutions. This paper builds on previous results of the authors regarding the minimum number of processors that is necessary to allow finding a feasible schedule for a given task set. As previous work was considering single-instance tasks, now the focus moves to periodic tasks, and the existing results are extended in such a way as to cover the new requirements. Also, an existing scheduling algorithm, which aims to combine the characteristics of the well-known EDF and LLF techniques, is being adapted for dealing with periodic tasks.