Mixed-criticality real-time systems, where tasks may be associated with different criticality and assurance levels, have attracted much attention in the recent past. In this paper, we consider partitioning-based multiprocessor scheduling of mixed-criticality real-time task sets. Guaranteeing feasibility in this setting is shown to be NP-Hard. With a focus on fixed-priority preemptive scheduling on each processor, we identify the two main aspects of the problem, namely the task allocation and priority assignment dimensions. For the task allocation dimension, we propose and compare bin-packing-inspired heuristics, based on offline task ordering according to utilization and criticality. For the priority assignment dimension, we compare the well-known Rate Monotonic priority assignment policy with Audsley's priority assignment algorithm. Through simulations, we also assess and discuss the relative importance of these two primary dimensions on the overall mixed-criticality feasibility problem for multiprocessor platforms.
Download Full PDF Version (Non-Commercial Use)