Validating real-time constraints in heterogeneous multiprocessor and distributed systems

Abstract

A heterogeneous system or distributed system typically contains different types of processors. Some jobs can execute on certain types of processors but not other types. Processors of different types may have different speeds. We present several worst-case upper bounds and efficient algorithms for determining the completion times of independent jobs that have arbitrary timing constraints and variable execution times and are scheduled on available processors dynamically in a priority-driven manner. Bounds and algorithms take into account the effects of nonmigratability, nonpreemptability and different levels of migratability.