What if we would degrade LO tasks in mixed-criticality systems?

Mixed-criticality (MC) systems [1] allow tasks of different importance (or criticality) to be consolidated into a single system. Consolidation facilitates resource sharing (even across criticality levels) and hence bears the potential to reduce the overall amount of resources needed. However, there is a common misconception that recurs in literature about Vestal’s model: the false believe that low criticality tasks are degraded to soft real-time or even best effort tasks. In this work, we not only wish to clarify this misconception but also ask ourselves what would happen if we degrade LO tasks. Revisiting Quality Assuring Scheduling (QAS) [2], [3], our goals are stochastic guarantees for LO completion in addition to and replacing the hard MC guarantees if LO tasks are soft real-time. In this WIP report, we focus on properties of dropped LO tasks (while keeping hard MC guarantees) such as: “What is the likelihood of lower criticality jobs being dropped because higher criticality jobs exceed their low WCET estimates?”, “What is the likelihood of dropped jobs to still make their deadline?”, and “What is the expected time / Q-percentile for dropped jobs to catch up with their execution?”. Part of our future work will be to extend these guarantees and to develop MC schedulers for a combination of hard and soft real-time tasks. Notice though, that the assumptions in this report still limit the applicability of our results. We indicate how we plan to relax them in the future. Most notably, we assume that jobs arrive in their synchronous arrival sequence and that executiontime distributions are known. The latter we plan to replace with confidence of WCET estimates.