Predictive modeling for job power consumption in HPC systems

Abstract

Power consumption is a critical aspect for next generation High Performance Computing systems: Supercomputers are expected to reach Exascale in 2023 but this will require a significant improvement in terms of energy efficiency. In this domain, power-capping can significant increase the final energy-efficiency by cutting cooling effort and worst-case design margins. A key aspect for an optimal implementation of power capping is the ability to estimate the power consumption of HPC applications before they run on the real system. In this paper we propose a Machine-Learning approach, based on the user and application resource request, to accurately predict the power consumption of typical supercomputer workloads. We demonstrate our method on real production workloads executed on the Eurora supercomputer hosted at CINECA computing center in Bologna and we provide useful insights to apply our technique in other installations.