Divisible load scheduling of real-time task on heterogeneous clusters

Abstract

Recent research in real-time scheduling had applied workload models from Divisible Load Theory (DLT) to real-time workloads. The resulting theory, referred to as Real-time Divisible Load Theory (RT-DLT), hold great promise for modeling an emergent class of massively parallel real-time workloads. However, the theory needs strong formal foundations before it can be widely used for the design and analysis of real-time systems. In this paper, we report our current findings on RT-DLT as part of our on-going research effort to develop such formal foundations. Specifically, we have extended an algorithm based on Linear Programming to schedule a divisible real-time workload upon heterogeneous clusters. A series of simulations have been conducted and a set of significant results are presented.