Towards tight Best- and Worst-Case Bounds with Heaviside Real-Time Analysis

Abstract:
In real-time scheduling theory, most of the literature focusses on worst-case response time analysis to predict a deadline violation. The extension of the analysis from singlecore processors to distributed real-time systems also requires the computation of the best-case response time to analyze end-to-end latencies. However, less work is published for bestcase analysis. On top of that, the approaches for best- and worst-case analysis differ, although, intuitively, they should have a symmetry. In this paper, we present for the first time a general response time analysis for fixed-priority schedulings where the best- and worst-case are derived as special cases of a time-based response time analysis. The time-based model enables us to prove the tightness of our bounds. In experiments, we compare a tight best- and worst-case response time analysis to a simulation of response times in the same model for an industrial task set and discuss the probability of the worst-case.