Algorithms for implementing elastic tasks on multiprocessor platforms: a comparative evaluation

Abstract

The elastic task model enables the adaptation of systems of recurrent real-time tasks under uncertain or potentially overloaded conditions. A range of permissible periods is specified for each task in this model; during run-time a period is selected for each task from the specified range of permissible periods to ensure schedulability in a manner that maximizes the quality of provided service. This model was originally defined for sequential tasks executing upon a preemptive uniprocessor platform; here we consider the implementation of sequential tasks upon multiprocessor platforms. We define algorithms for scheduling sequential elastic tasks under the global and partitioned paradigms of multiprocessor scheduling for both dynamic and static-priority tasks, and we provide an extensive simulation-based comparison of the different approaches.