Computational Science is often considered a third mode of science, complementing theory and observation. They have become indispensable in science, meaning that large supercomputing centres are used daily for simulations and data analysis. However, this growing demand for computing power has a downside: enormous energy consumption and associated CO₂ emissions from cooling systems and hardware. It was estimated that in 2020 the ICT sector was responsible for 1.8% - 2.8% of global greenhouse gas emissions. Recent developments in the field of AI will cause energy consumption to rise even further.
Reducing energy consumption
The ECO-COMPASS project aims to enable researchers to make an informed trade-off between speed, accuracy and energy consumption. The project is focusing on three key pillars:
- Validating existing methods by using the KernelTuner software tool in case studies to make algorithms more efficient in computational imaging, computational fluid dynamics and machine learning.
- Developing new knowledge by benchmarking the selected case studies on new, energy-efficient hardware.
- Knowledge sharing and awareness-raising by widely disseminating the knowledge gained through tutorials and training courses, so that this knowledge becomes accessible to as wide an audience as possible.
The ECO-COMPASS guidelines and tutorials will be made public and disseminated via partner networks such as SURFsara, NWO-I and the Netherlands eScience Center. In this way, the consortium aims to set a new standard for sustainable research in the Netherlands.
