Exploring Busy Period for Worst-Case Deadline Failure Probability Analysis

Abstract

Busy period is a fundamental concept in classical deterministic real-time scheduling analysis. In this deterministic context, only one busy period - which starts at the critical instant - needs to be considered, which identifies the worst-case scenario and thus paves the way for the development of efficient and safe analysis techniques. However, a recent work has revealed that, in the context of \textit{probabilistic} real-time scheduling analysis, only considering critical instant is not safe. In this paper, we address this gap by systematically analyzing deadline miss probabilities across varying busy period starting points. We propose a novel method of Worst-Case Deadline Failure Probability (WCDFP) for probabilistic fixed-priority preemptive scheduling. Experimental results demonstrate significant improvements over state-of-the-art methods achieved by our proposed method.