Cluster wide energy and power optimisation software to maximise the efficiency of your HPC cluster
Energy is a critical resource for data centres and the planet. HPC installations need to provide maximum results from the power they consume. EAS offers solutions for data centre energy optimization to reduce cost and be eco-responsible through the installation, configuration, support and services of the Energy Aware Runtime (EAR) open-source software.
OCF is proud to partner with Energy Aware Solutions to offer this exciting energy optimization solution to its customers in the UK and beyond.
EAR is an energy management software for the efficient energy control of heterogeneous clusters.
EAR dynamically and automatically measures job power and performance metrics and then optimise the node settings to run as energy efficiently as possible without any user intervention. EAR reports all the job metrics tied to the scheduler jobids to make it easy the application analysis. For example EAR4.0 includes a Slurm plugin for node and job control (SPANK).
EAR provides power/energy control, accounting, monitoring and optimization of both the applications running on the cluster and the overall global cluster in a fully transparent way to the users.
EAR 4.0 released in August 2021 provides energy and power capping policies on Intel/AMD processors and NVIDIA GPUs. With this version, EAR has demonstrated on a large set of applications energy savings of around 10% on homogenous and hybrid clusters with Intel CPUs and NVIDIA GPUs.
EAR is proven at scale, optimising the energy consumption of the 3MW SuperMUC HPC facility at LRZ. However, benefits can be experienced with clusters of any size. It supports a monitoring mode where it analyses the energy characteristics of the cluster in a dormant fashion which can be a useful first step is seeing if EAR would work for your workload.