Towards Generalizing "Big.Little" for Energy Proportional HPC and Cloud Infrastructures

Abstract

Reducing energy consumption is part of the main concerns in cloud and HPC environments. Today servers energy consumption is far from ideal, mostly because it remains very high even with low usage state. An energy consumption proportional to the server load would bring important savings in terms of electricity consumption and then financial costs for a data enter infrastructure. In this paper, we propose a platform composed of heterogeneous architectures to achieve proportional computing goal. We select low power ARM processor for a light load, and a range of regular x86 servers when performance is required. We propose a comparative study of benchmark execution in order to find the best configuration depending on the current load and show the effective results in terms of energy proportionality.