Mixed Criticality Systems with Varying Context Switch Costs

In mixed criticality systems, it is vital to ensurethat there is sufficient separation between tasks of LO- andHI-criticality applications, so that the behavior or mis-behaviorof the former cannot affect the functional or timing correctnessof the latter. To ensure appropriate spatial isolation, thememory address spaces and cache use by LO- and HI-criticalitytasks must be distinct. A consequence of this separation is thatthe cost of switching between tasks of the same criticality canbe small, whereas the cost of context switching between tasksof different criticality levels can be much larger. In this paper,we focus on integrating the differing context switch costs intofixed priority preemptive scheduling, and the two mixedcriticality scheduling schemes based on it: SMC and AMC. Wederive simple, refined, and multi-set analyses for each scheme.Further, we show that the refined and multi-set analyses arenot compatible with Audsley’s Optimal Priority Assignmentalgorithm, we therefore propose a heuristic priority assignmentpolicy aimed at reducing the number of high cost contextswitches. Our evaluation is grounded in measurements ofcontext switch times (save and restore costs) from a prototypeimplementation of an explicitly managed cache on an FPGA.The evaluation shows the effectiveness of the derived analysesand the proposed priority assignment policy.