Building an AWS EC2 Carbon Emissions Dataset
summary
Builds an EC2 carbon dataset and a public calculator using on-instance RAPL readings accessed via turbostress (stress-ng + turbostat) to profile Intel bare‑metal CPU+memory power, then scales those profiles to instance sizes. Finds peaks around 900 W on c5 metal under full load, adds a heuristic uplift for unmeasured components, and models manufacturing as a naive 1 ton manufacturing baseline per minimal server with a 4‑year lifespan to get hourly embodied CO2. Limits: no AMD/Graviton coverage and virtualization assumptions.
AI-written summary · about
chapters (6)
- 0:00Start
- 1:16Why Did We Do this Build this Data Set for Ec2
- 6:27Build Consumption Profiles
- 10:14How Virtual Machines Are Defined by Aws
- 14:00Naive Model To Estimate the Emissions from Manufacturing
- 17:16Cloud Carbon Footprint Project
related talks
44mCloud computing and carbon emissions
50mResponsible computing: how about a survey of measurement tools?
22mICT & cloud footprint data at scale: [...]
41mWriting Greener Software Even When You Are Stuck On-Prem
31mTrack your code's CO2 emissions with Code Carbon
20mBuilding GenAI That Doesn't Cost the Earth
25mBuilding a Sustainable Future
42mBuilding a green supercomputer in the cloud