Building a green supercomputer in the cloud
summary
A cloud HPC cluster runs Vestas wind turbine simulations at up to 400,000 cores and handles multiple petabytes of data, built in eight months on Azure with .NET and ARM-friendly instances. Time-shifting jobs cuts carbon 12–16%; location-shifting within the EU yields ~30–90% savings by moving work from West Europe to Sweden Central. The system uses ElectricityMap and the Carbon Aware SDK plus spot instances and a custom Carbon Economy SDK to balance emissions against cost, and streams outputs back in parallel to minimize delays.
AI-written summary · about
chapters (15)
- 0:00Introduction to green supercomputing
- 1:52Vestas wind turbine challenge
- 5:48Green software foundation
- 8:24Energy efficiency strategies
- 10:11Carbon awareness explained
- 17:30CO₂ equivalence and intensity
- 22:08Time shifting workloads
- 27:24Location shifting workloads
- 35:24Results of time/location shifting
- 36:05Tools for software carbon intensity
- 42:12Calculating software carbon intensity
- 45:01Carbon economy SDK
- 50:20Solution for Vestas
- 53:09Microsoft carbon tools
- 54:37Q&A
related talks
42mBuilding a green supercomputer in the cloud
44mCloud computing and carbon emissions
25mBuilding a Sustainable Future
41mToward a carbon-aware cloud
41mWriting Greener Software Even When You Are Stuck On-Prem
42mUndertaking a GreenOps Approach in the Public Cloud
20mBuilding GenAI That Doesn't Cost the Earth
33mSmart Green Computing Cloud Native Operations