Accelerating Science, Decelerating Carbon: Sustainable Computing...
summary
CERN builds a metrics-driven framework that maps per-pod energy and cost to CO2 using OpenCost and Kepler with Prometheus exporters, collecting node/pod/container/process power and cloud billing so workloads get cost, joule and carbon attribution. It runs on on‑prem and Oracle Cloud, profiles real HEP workloads (NGT, ATLAS-scale 5,000-node migrations), produces sustainability scores (cost per joule, performance per watt) and plans dynamic scheduling and hybrid-cloud placement driven by those scores.
AI-written summary · about
chapters (14)
- 0:00CERN project introduction
- 3:51Carbon aware computing at CERN
- 5:14Environmental objectives and metrics
- 6:18Data pipeline with OpenCost and Kepler
- 8:48How Kepler measures energy
- 11:16Testing tools at CERN and Oracle
- 11:56Energy data and carbon intensity
- 13:01Next Generation Triggers context
- 14:44Monitoring dashboards for metrics
- 16:08Energy efficiency per pod and node
- 17:14Carbon emissions examples
- 18:19Sustainability scoring metrics
- 19:21Phase three scheduling policies
- 20:23Q&A
related talks
36mCarbon-Aware Computing: Measuring and Reducing the Carbon Intensity of Software
18mKeynote: Sustainability Chronicles: Innovate Through Green Technology with...
42mBuilding a green supercomputer in the cloud
20mMonitoring Energy Consumption and Building Energy Efficient Systems the Cloud Native Way
20mEnergy Efficient Placement of Edge Workloads
55mToward a carbon-aware Cloud
17mKepler: Powering up the efficiency of your cluster
28mExploring Kepler's Next Chapter: Achieving Cloud Native Sustainability With MCP Integration