Elastic Scheduling for Fixed-Priority Constrained-Deadline Tasks

Elastic scheduling provides a model for systems in which individual task utilizations can adapt to guarantee schedulability despite limited resources. Each task is characterized by a range of acceptable utilizations and an "elastic constant" representing its flexibility to reduce or "compress" its utilization from the desired maximum. Utilization compression is realized by either extending task periods or reducing workloads. This paper extends the model to address period compression for fixed-priority constrained-deadline task systems scheduled on a uniprocessor. We propose two approximate algorithms and one optimal algorithm for determining compression under the model. We then compare the execution times and accuracies of all three, demonstrating that even for large task sets, online compression can be performed feasibly on low-powered embedded systems.