Global Rate-Monotonic Scheduling with Priority Promotion

In this paper, we consider a multicore real-time scheduling algorithm incorporating benefits of both fixed-priority and dynamic-priority disciplines. Specifically, the algorithm first assigns globally-effective priorities to real-time tasks statically, based on the well-known Rate-Monotonic scheduling policy. It may however change the task priorities at runtime, only when the tasks reach the zero-laxity condition, where no slack remains until the deadline, to avoid timing violations as much as possible. Implementation simplicity and response time predictability are therefore inherited from the fixed-priority discipline, while minimal dynamicpriorities are exploited, if necessary, to maintain the system to be schedulable as much as possible. We also provide a schedulability analysis and derive a schedulability test for the algorithm. Our evaluation then demonstrates that the algorithm outperforms the existing global fixed-priority scheduling algorithms in terms of schedulability.