Towards the design of fault-tolerant mixed-criticality systems on multicores.

Mixed-criticality is a significant recent trend in the embedded system industry, where common computing platforms are utilized to host functionalities of varying criticality levels. To date, most scheduling techniques have focused on the timing aspect of this problem, while functional safety (i.e. fault-tolerance) is often neglected. This paper presents design methodologies to guarantee both safety and schedulability for real-time mixed-criticality systems on identical multicores. Assuming hardware/software transient errors, we model safety requirements on different criticality levels explicitly according to safety standards; based on this, we further propose fault-tolerant mixed-criticality scheduling techniques with task replication and re-execution to enhance system safety. To cope with runtime urgencies where critical tasks do not succeed after a certain number of trials, our techniques can perform system reconfigurations (task killing or service degradation) in those situations to reallocate system resources to the critical tasks. Due to explicit modeling of safety, we can quantify the impact of task killing and service degradation on system feasibility (safety and schedulability), enabling a rigorous design. To this end, we derive analysis techniques when reconfigurations are triggered either globally (synchronously) on all cores or locally (asynchronously) on each core. To our best knowledge, this is the first work on fault-tolerant mixed-criticality scheduling on multicores, matching theoretical insights with industrial safety standards. Our proposed techniques are validated with an industrial application and extensive simulations.