Energy-aware heuristics for scheduling parallel applications on high performance computing platforms

Abstract

The massive demand for running parallel applications on distributed systems has led to an upsurge in the system power consumption. These systems often consist of thousands or millions of cores, storage disks, interconnection devices and other power-hungry components. To address this power consumption problem, we propose two energy-aware scheduling algorithms, namely, Energy-Efficiency with Duplication (EED) and Energy-Efficiency with Non Duplication (EEND). Both algorithms, in contrast to their counterparts in the literature, strive to make a balance across the energy consumption, the schedule length, and the number of processors used. Synthetic benchmarks and real-world applications are used to evaluate the performance of our algorithms. A comparison is made with the Traditional Duplication-based Scheduling algorithm (TDS), the Energy Aware Duplication algorithm (EAD), and the Performance-Energy Balanced Duplication algorithm (PEBD). We have also devised a novel objective function that aids in assessing the performance of different algorithms in terms of various performance metrics. Extensive experimental results conducted as a part of this work show that the performance of our algorithms is promising in terms of energy consumption, processing elements used and the schedule length for communication intensive parallel applications.