Deterministic and Probabilistic Schedulability Analysis for Real-Time Systems

Abstract

Traditional deterministic schedulability analysis provides a binary guarantee—a task set is either schedulable or not—based on fixed worst-case execution times (WCET). While safe, this approach often leads to pessimistic and resource-inefficient system designs, as it cannot account for the runtime variations inherent in complex computing platforms. This paper introduces a deterministic schedulability analysis and a new paradigm for schedulability analysis founded on probabilistic principles. Our approach generalizes the traditional task model by explicitly incorporating execution time variations. Rather than relying on static WCETs, the proposed model captures this variability using probabilistic distributions, clustering, and adaptive estimation techniques. This shift enables a more realistic and flexible assessment of system timing.