REALISM: Reliability-aware energy management in multi-level mixed-criticality systems with service level degradation

Abstract

Mixed-criticality embedded systems, as the next-generation of safety-critical systems, are increasingly employed in the industry due to consolidating functionalities with varying criticality levels onto the same computing platform. Technology scaling, battery-supplied design, and heavy computation in mixed-criticality systems necessitate employing energy management techniques. Due to the degrading effects of these techniques on the system's reliability, minimizing energy consumption and meeting tasks’ timing constraints without sacrificing the reliability requirement is a vital challenge in designing mixed-criticality systems. In this paper, we propose REALISM; a novel Reliability- and Energy-Aware task mapping and scheduling approach in multi-core mixed-criticality systems with three levels of criticality. To guarantee the desired QoS of tasks from lower levels of criticality, our approach exploits the degraded WCET for these imprecise mixed-criticality tasks in critical modes. In design time, tasks are mapped to cores based on dynamic switching between different system utilizations in different modes. In each core, task instances are scheduled according to their virtual deadlines, which are calculated based on the system's overrun tolerance limit (TL) and the desired QoS, by time reservation. Then our heuristic DVFS techniques are applied in order to reduce system energy consumption. Our simulation results show that the REALISM provides on average ~41.5% (up to 52.7%) energy savings compared to the approach without energy management with 100% QoS degradation in lower levels of criticality. Also, with guaranteeing the system's desired QoS, REALISM provides on average ~38% (up to 46.2%) energy savings compared to the approach without energy management.