Integrating Proactive Mode Changes in Mixed Criticality Systems

In this work, we propose to integrate prediction algorithms to the scheduling of mode changes under the Earliest-Deadline-First and Fixed-priority scheduling in mixed-criticality real-time systems. The method proactively schedules a mode change in the system based on state variables such as laxity, to the percentage difference in the temporal distance between the completion time of the instance of a task and its respective deadline, by the deadline (D) stipulated for the task, in order to minimize deadline misses. The simulation model was validated against an analytical model prior to the logical integration of the Kalman-based prediction algorithm. Two study cases were presented, one covering earliest-deadline first and the other the fixed-priority scheduling approach. The results showed the gains in the adoption of the prediction approach for both scheduling paradigms by presenting a significant reduction of the number of missed deadlines for low-criticality tasks.