Energy-aware scheduling for parallel evolutionary algorithms in heterogeneous architectures

The availability of mechanisms such as dynamic voltage and frequency scaling (DVFS) and heterogeneous architectures including processors with different power consumption profiles allow scheduling algorithms aware of both runtime and energy. In this paper, we propose and evaluate a scheduling strategy that takes into account the relative weights of the workloads and the frequencies and voltages of the different processing cores in a given heterogeneous parallel architecture either to save energy without increasing the running time or to reach a trade-off among time and energy. The parallel algorithms considered to evaluate the proposed scheduling procedure are master-worker evolutionary algorithms whose fitness functions demand high computing times and distribute the fitness evaluation of the individuals among the available cores. As many useful bioinformatics and data mining applications present this profile, the proposed energy-aware scheduling approach could be frequently applied. The experimental results obtained by simulation show relevant energy savings, with values depending on the characteristics of the heterogeneous architecture and on the workload profiles