Provided by: University of Calgary
Date Added: Nov 2011
Optimal multiprocessor real-time scheduling is challenging. Several solutions have recently been presented, most based on Periodic-Preemptive-InDependent (PPID) tasks with implicit deadlines. The authors address a generalization of this with the goal of finding an efficient valid schedule (i.e., meeting all deadlines). Multiprocessor scheduling is often achieved via partitioned approaches in which tasks are statically assigned to processors, guaranteeing only 50% utilization in the worst case. Global approaches can achieve full utilization by migrating tasks between processors, at the cost of increased runtime overhead.