A probabilistic scheduling framework for mixed-criticality systems

We propose a probabilistic scheduling framework for the design and development of mixed-criticality systems, i.e., where tasks with different levels of criticality need to be scheduled on a shared resource. Whereas highly critical tasks normally require hard real-time guarantees, less or non-critical ones may be degraded or even temporarily discarded at runtime. We hence propose giving probabilistic {instead of deterministic) real-time guarantees on low-criticality tasks. This simplifies the analysis and reduces conservativeness on the one hand. On the other hand, probabilistic guarantees can be tuned by the designer to reach a desired level of assurance. We illustrate these and other benefits of our framework based on extensive simulations.