Energy Consumption Studies of WRF Executions with the LIMITLESS Monitor

Abstract

We present a study of the performance of the Weather Research

and Forecasting [WRF] code under several hardware con gurations

in an HPC environment. The WRF code is a standard code for

weather prediction, used in several elds of science and industry. The

metrics used in this case are the execution time of the run and the energy

consumption of the simulation obtained with the LIMITLESS monitor,

which is the main novelty of this work. With these results it is possible

to quantify the energy savings of WRF run con gurations, which include

variations in the number of computing nodes and in the number of processes

per node. It is found out that a slight increase in the computing

time can drive to a noticeable reduction in the energy consumption of

the cluster.