/images/overview.png
/images/logo.png

Energy efficient HPC

The energy consumption of HPC data centers is a decisive factor in the procurement and operation of the systems. EE-HPC achieves a more efficient energy use of HPC systems by targeted job-specific control and optimization of the hardware configuration as well as of settings of the runtime environments.

Project Structure

The aim of the project is the automated optimization of the energy efficiency of HPC systems. An innovative monitoring system is to contribute to reducing energy consumption while simultaneously increasing computing performance. This goal is to be achieved by new software-based control mechanisms of system parameters. The adjustment of system parameters, such as the utilization of computing nodes, is to take place automatically. A monitoring software coupled with a novel user interface shall provide the user with a transparent platform to also decide on the energy efficiency part of the computing load. This holistic approach ensures flexible and broad use for a wide range of applications.

Work package 1

Node agent

WP1 integrates the ClusterCockpit node agent and LIKWID into the GEOPM node agent. Furthermore, this work package deals with the acquisition of runtime parameters as well as program control and adaptation of system parameters.

Work package 2

User interface

WP2 provides the necessary monitoring infrastructure including data backends. Furthermore, the web-based graphical user interface is implemented here.

Work package 3

Runtime systems and programming models

Development and implementation of interfaces and components for the interaction of the analysis and monitoring software with the runtime environments of OpenMP and MPI These components provide input to the tuners and trigger re-evaluations.

Work package 4

Analysis and optimisation

This work package develops the actual optimization layer as an extension of GEOPM.

Work package 5

Integration and evaluation

In this work package, all developed software components will be installed at the participating HPC centers, (as far as possible) taken over into user operation and maintained. The approaches will be tested in detail - in real operation - and the effectiveness of the parameter tuning will be validated on the basis of production codes.

Project partners

Funding

/images/bmbf.jpg

Project duration: 09/2022 - 08/2025

EE-HPC is funded by BMBF Germany under the funding program Hoch- und Höchstleistungsrechnen für das digitale Zeitalter 2021-2024 – Forschung und Investitionen zum High-Performance Computing, Call Richtlinie zur Förderung von Verbundprojekten auf dem Gebiet des energieeffizienten High-Performance Computings (GreenHPC)

Grant 16ME0583K