From Heterogeneous Task Scheduling to Heterogeneous Mixed Data and Task Parallel Scheduling
Frédéric Suter, Henri Casanova, Frédéric Desprez, Vincent Boudet · 2003
Mixed-parallelism, the combination of data- and task-parallelism, is a powerful way of increasing the scalability of entire classes of parallel applications. Exploiting both types of parallelism simultaneously makes it possible to deploy these applications on platforms comprising multiple compute clusters, which have become increasingly popular in the last decade. However, high performance application executions are only possible if effective scheduling strategies are available. While multi-cluster platforms are predominantly heterogeneous, previous work on mixed-parallel application scheduling targets only homogeneous platforms. In this paper we develop a method for extending existing scheduling algorithms for task-parallel applications on heterogeneous platforms to the mixed-parallel case. After detailing the foundations of our method and our assumptions, we present a case study in which we generate a mixed-parallel version of the popular HEFT scheduling algorithm, which we evaluate with an extensive set of simulation experiments.