EECliPs: Energy-efficient climate simulations

Climate research based on numerical modells is intended to contribute to a better understanding of climate change and its effects, and increasingly to the assessment of adaptation strategies. State-of-the-art high performance computers are indispensable for the calcualtion of these models, but are very energy-intensive to operate. The EECliPs project is investigating how energy-efficient execution of climate simulations can be achieved through a suitable co-design of HPC hardware and software.

On November 15, 2022, during their kickoff meeting, the partners discussed the first concrete steps to evaluate and implement the various approaches. As the coordinator, the DKRZ will take over the integration into its infrastructure and the implementation of the climate simulations. The ZIH of the TU Dresden contributes the necessary expertise in tools for the analysis and energy measurement of HPC applications and ParTec together with Atos takes over the hardware-side investigations. The goal is an optimal load balancing of the climate and weather model ICON on a heterogeneous proof-of-concept system consisting of different current architectures. The trend towards specialized architectures in the HPC environment is exploited, so that the energy consumption of a simulation is reduced without significantly increasing the execution time.

EE-HPC: Energy optimization of data centers

The project "Open source solutions for monitoring and system settings for energy-optimized computing centers" (EE-HPC) is funded by the BMBF. The priject is coordinated by the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), further partners are  DKRZ,  RWTH Aachen, HLRS, as well as Hewlett-Packard and Intel Germany as industrial partners. They met for a kick-off meeting from October 27th to 28th, 2022 in Erlangen.

The aim of the project is the automated optimization of the energy efficiency of HPC systems. By means of job-specific control and optimization of the hardware configuration and the runtime environments (OpenMP and MPI), clusters are intended to achieve more efficient energy use by reducing power consumption while at the same time maximizing throughput. A system-wide, job-specific framework for performance and energy monitoring, which is based on ClusterCockpit, serves as the basis. Analytical modelling, machine learning and empirical methods are combined to determine the parameters. The overall solution should be usable for a wide variety of applications and will be tested on the productive systems of all project partners. At DKRZ, the environment is used as an example to test applications in the field of climate research on different HPC architectures and to determine the most energy-efficient solution.