Thermal-Aware Power Capping Allocation Model for High Performance Computing Systems

High-performance computing (HPC) systems are large computing infrastructures, which consume massive amount of power during their operation. Power capping is a feature introduced in modern processor architecture to control application performance running on compute nodes. In this paper, we exploit power capping capability in the processors to develop a thermal-aware energy-efficient model for HPC systems. Our model optimizes energy consumption of HPC applications, while ensures processor temperature remains within a limit. We execute various HPC applications and measure different characteristics of execution (e.g., power, performance, temperature). Based on real-life measurements, we demonstrate that our proposed model is effective on achieving thermal-aware energy-efficiency for HPC systems.