EDF Scheduling of Real-Time Tasks on Multiple Cores: Adaptive Partitioning vs. Global Scheduling

This paper presents a novel migration algorithm for real-time tasks on multicore systems, based on the idea of migrating tasks only when strictly needed to respect their temporal constraints and a combination of this new algorithm with EDF scheduling. This new "adaptive migration" algorithm is evaluated through an extensive set of simulations showing good performance when compared with global or partitioned EDF: our results highlight that it provides a worst-case utilisation bound similar to partitioned EDF for hard real-time tasks and an empirical tardiness bound (like global EDF) for soft real-time tasks. Therefore, the proposed scheduler is effective for dealing with both hard and soft real-time workloads.