Approximation scheme for real-time tasks under fixed-priority scheduling with deferred preemption

Abstract

We consider real-time tasks under fixed-priority scheduling with deferred preemption (FPDS) and arbitrary deadlines. Exact response time analysis is known but exhibits to high computational cost. Effective sufficient test has been proposed under certain conditions and still subjects to pseudo-polynomial complexity. Some applications require efficiently computed response time bounds. In this paper, we provide a novel exact processor demand analysis. Based on this, we define a parametric polynomial time algorithm that both analyzes the approximate feasibility and computes the response time upper bounds. Using the resource augmentation technique, we obtain a performance guarantee that allows defining a compromise between the response-time bound and processor capacity requirements. Finally, the algorithm average behavior is analyzed through numerical experiments.