SMART-EDF: An EDF based semi-partitioned energy-aware multicore scheduler for real-time systems

In real-time systems, schedulers have to perform tasks with a specific degree of urgency; otherwise, it may lead to significant damage. Additionally, modern systems also want to focus on other design criteria like providing high system utilization, executing tasks with a low energy budget etc., making the job of a scheduler more challenging in such systems. EDF is one of the first optimal schedulers designed for single-core real-time systems, and since then, it has remained in the focus of the research fraternity. Unfortunately, it is not suitable for multicore platforms. In this paper, we have designed a semi-partitioned energy-aware scheduler for real-time systems based on EDF. Initially, the proposed scheduler uses the strategy of deadline partitioning to make the algorithm semi-partitioned. Then, it applies EDF to schedule the tasks on available cores. Next, it carefully schedules tasks that require migration across multiple cores. At last, it scales the frequency of individual cores based on the workload assignment of previous steps. Our experimental analysis shows that the proposed scheduler can provide high resource utilization (unlike traditional multicore schedulers) and reduce dynamic energy consumption (up to 8.94% on an average) of the systems.