Mixed-Criticality Scheduling on Multiprocessors with Service Guarantees

Mixed-criticality (MC) systems are composed of tasks with varying criticality co-hosted on a single shared platform. In conventional MC systems, upon criticality change, the lower criticality tasks are penalized to guarantee resources for the higher criticality ones. However, in practice, penalizing lower criticality tasks have adverse effects and hence, the system is often under-utilized. In this paper, we consider the problem of reservation-based scheduling of mixed-criticality systems on a homogeneous multiprocessor platform to guarantee full service to the lower criticality tasks when one of the processors switches to the critical state. We explore the semi-partitioned scheduling model for dual-criticality systems in which the low criticality tasks executing on a processor are migrated to another processor upon mode switch to improve the service offered to them in the high criticality mode. We present the scheduling strategy of the proposed algorithm and derive its utilization bound. To evaluate the proposed algorithm, we use randomly generated task sets to compare the schedulability performance of the algorithm with the existing algorithms. Our results show that the proposed algorithm improves both schedulability and low criticality support when compared to existing algorithms for implicit-deadline task systems.