Energy efficiency management in computational grids through energy-aware scheduling

Energy consumption in High Performance Computing (HPC) has become an important issue in the past few years. The performance gain obtained by these environments is matched by a proportional increase of energy use. Example of such environments are computational grids, which are used in several academic and enterprise projects. Given this scenario, researchers have been trying to reduce the energy consumption while minimizing performance loss at the same time. This work proposes the use of energy-aware scheduling for energy efficiency management in computational grids. Our solution exploits the main existing approaches in the literature to reduce energy consumption in HPC environments: management of idle resources and energy-aware scheduling algorithms. We evaluate our proposed approach in a simulation environment and the algorithm was compared to other five traditional scheduling algorithms that do not consider energy features. Results show an energy reduction of up to 182.90% combined with a performance loss up to 27.78% in the best cases.