State-based scheduling analysis for distributed real-time systems

Abstract

The amount of system functions realized by software drastically increased in recent years. Software tasks of safety-critical systems like those in the automotive domain have to work in a timely manner. In such systems not only ordering of events but also timing properties like end-to-end deadlines are relevant for correctness and performance. Unfortunately, due to various inter-dependencies between software tasks the analysis of such properties becomes very complex. The state-of-the-art analysis approach considers only stateless system behaviors and relies on critical instances leading to very pessimistic results. Considering task inter-dependencies would result in more accurate results, though it negatively affects the scalability of the analysis. Our approach for scheduling analysis combines analytical and model checking methods. We consider the full state space of a system, where all interleavings and task dependencies are preserved. The state space is build in a compositional manner enabling a more scalable technique. For this, we introduce operations on the state spaces of resources, allowing the abstraction of irrelevant parts and the composition of state spaces. Based on the state space of each resource response times are determined, and timing and safety properties can be verified by means of reachability checks. The approach is demonstrated based on an example scenario.