Predicting the Energy Consumption of MPI Applications at Scale Using a Single Node

International audienceMonitoring and assessing the energy efficiency of supercomputers and data centers is crucial in order to limit their energy consumption. Applications from the domain of High Performance Computing (HPC) consume a significant fraction of the overall energy consumed by HPC centers. Simulation is a popular approach for studying the behavior of HPC applications in a variety of scenarios, and it would therefore be advantageous to study the energy consumption of HPC applications in a cost-efficient, controllable, and also reproducible simulation environment. Unfortunately, simulators of HPC applications (e.g., MPI applications) typically lack the capability of predicting the energy consumption, in particular when platforms consist of multi-core nodes. In this work, we aim to accurately predict the energy consumption of MPI applications via simulation. First, we introduce the models required for meaningful simulations, which are the computation model, the communication model, and the energy model of the target platform. Second, we show that by carefully calibrating these models on a single node, the predicted energy consumption of HPC applications at a larger scale is very close (within a few percents) to real experiments. We show how to integrate such models into the SimGrid simulation toolkit. In order to obtain good execution time predictions on multi-core architectures, we show it is essential to correctly account for memory effects in simulation. The proposed simulator is validated through an extensive set of experiments with common HPC benchmarks. Finally, we show the simulator can be used for studying applications at scale, which allows to save both experimental time and resources compared to real experiments