Challenge:
As sustainability became a central goal for SURF’s national research infrastructure, they needed a way to reduce the energy usage of their new Snellius supercomputer without compromising performance. Hybrid workloads combining CPUs and GPUs posed unique monitoring and optimization challenges, especially for scientific applications in computational chemistry and machine learning.
Solution:
SURF partnered with the EAR team to deploy EAR across the hybrid Snellius system. EAR enabled job-level energy monitoring, dynamic energy policy testing, and real-time performance-per-watt insights, all essential for guiding future sustainability policies.
Results:
EAR has provided actionable data for understanding workload-level energy consumption, helping SURF develop smarter energy strategies. The collaboration opened the door to new research into software-centric energy optimization approaches and supports SURF’s long-term sustainability roadmap.
About SURF:
SURF is the collaborative organization for IT in Dutch education and research. It provides cutting-edge computing and data infrastructure to support innovation in science and higher education across the Netherlands.
Using EAR since:
- May 2022 on Snellius
- November 2023 on Snellius 2
Systems:
- 576 AMD Rome CPU nodes
- 36 Intel CPU / 4× NVIDIA A100 GPU nodes
- 720 AMD Genoa CPU nodes