Skip to content

Green HPC: Sustainable Use of High Performance Computing

Green HPC refers to using high-performance computing systems in a carbon-efficient way, aiming to minimize environmental impact. Every user of Kelvin2 can contribute to reducing emissions by following sustainable computing practices.

Best Practices for Carbon-Efficient HPC Use

1. Run Only Meaningful Jobs

Before submitting any job, ensure it will generate valuable results. Avoid running calculations without a clear understanding of their purpose or relevance to your research.

2. Test Before Scaling

Always test input files and job scripts using small or short jobs before scaling up.

  • Use the k2-sandbox queue for initial testing.

  • This reduces the number of failed or cancelled jobs—preventing unnecessary carbon emissions.

3. Optimize Job Design

Consider whether your project goals can be achieved with fewer jobs. For instance:

  • Refine sampling strategies when exploring parameter spaces.

  • Use more efficient algorithms where possible.

Sustainability First: Avoiding Emissions Is Best

As with many areas of sustainability, the most effective approach is carbon abatement—not emitting carbon in the first place. Efficient HPC use reduces the need for offsetting emissions after they occur.

Learn More and Broaden Your Impact

For more guidance on reducing the carbon footprint of your digital research practices, explore the Digital Humanities Climate Coalition (DHCC) Toolkit

For AI workloads, see the “Maximal Computing” section.

  • The toolkit also offers advice on sustainable daily digital habits.

  • A holistic approach to sustainability—beyond just HPC—is crucial.

Institutional Commitment: QUB and Sustainability

Queen’s University Belfast promotes sustainability in research through its Concordat for Environmental Sustainability in Research and Innovation. Learn more about these initiatives here.

Take Action

By adopting these practices and thinking critically about your computational impact, you can contribute meaningfully to reducing carbon emissions in research.

Let’s make every computation count.