P51 - Optimization of Energy Efficiency of an HPC Cluster: On Metrics, Monitoring and Digital Twins

Abstract

In-depth monitoring together with a digital twin model allows to manage HPC clusters so that they can react to different external events. We shall use this information to improve operation of the cluster with respect to energy usage and cost of electricity, in order to maintain operation under reduced energy availability or to decrease the cluster's CO2 footprint. Different additional external data sources together with system monitoring and modelling allow to adjust the HPC cluster to meet these goals.