Carbon Emissions of Cloud Computing
Climate Goals of Cloud Providers
Amazon Web Services
Climate goal: Net-zero carbon by 2040
Microsoft Azure
Climate goal: Carbon negative by 2030, including scope 3 emissions
Microsoft Azure - Sustainability
Google Cloud
Climate goal: Energy supply CO2 free by 2030
Carbon Emissions by Region
Carbon Intensity of Data Centers (Heatmap) by Climatiq
Popular LinkedIn-Post by Mark Butcher comparing the carbon intensity of London, South Scotland and Dublin: https://www.linkedin.com/posts/markbutcher_sustainability-scope2-activity-7099703210608586752-nNrK?utm_source=share&utm_medium=member_desktop
Tools for Estimation, Measurement & Analytics
Estimate carbon emissions of cloud applications
Metrics
see also Data Center Sustainability Metrics
Carbon usage effectiveness (CUE)
- Indicates the carbon efficiency of a data center
- Metric is calculated by dividing the carbon emissions (kilograms of CO2e / year) generated by the data center by the energy used by its IT equipment.
Carbon-free energy percentage (CFE%)
- Represent the average percentage of carbon-free energy consumed by a data center in a particular location on an hourly basis.
- Good metric to be able to see how long my applications will run on carbon-free energy.
- Invented by Google in 2020. Published online: Carbon free energy for Google Cloud regions
- Google considers a location as "low carbon" if the CFE% value is greater or equal than 75 %.
Grid carbon intensity (gCO2eq/kWh)
- Indicates the average operational emissions per unit of energy from the grid for a specific data center.
- Good metric to compare the different regions in terms of Energy Carbon Intensity.
- Google uses the data from Electricity Maps. Google considers a location as "low carbon" if the grid carbon intensity is at max 200 gCO2eq/kWh.
GHG emissions including scope 3
- See GHG Protocol & Accounting of GHG Emissions from Digital Services & Companies
- Include the emissions that were caused by producing the hardware components, etc.
- Azure wants to include all their scope 3 emissions in their carbon accounting. See Microsoft Azure - Sustainability.
- Meta has been growing too fast to achieve its goal of being net zero in 2030. 99 % of the emissions are coming from their supply chain and ecosystem (Scope 3). Source: Meta report shows the company causes far more emissions than it can cover with renewable energy - DCD
Energy Intensity & Migration Paths
Cloud regions have different Energy Carbon Intensities. Therefore you can reduce the carbon emissions by moving to a different cloud region.
Cloud Carbon Explorer (Uptime Institute)
https://uptimeinstitute.com/resources/tools/cloud-carbon-explorer
The Cloud Carbon Explorer maps below show potential cross-region workload migration paths for AWS, Google Cloud and Microsoft Azure. These workload migration paths can reduce carbon footprint without significantly impacting user experience and, in some cases, reduce cost. Users can use the tool to explore suitable compromises of latency, cost and carbon for each application before pursuing their own more detailed assessments.
Google Cloud Regions
Carbon free energy for Google Cloud regions
Cloud Storage
Storage providers for remote backups#Storj
Storj claims edge in reducing carbon footprint – Blocks and Files
This claim rests upon two pillars. Firstly, Storj places its data on spare disk drive capacity in existing datacenters, capacity that is already powered and spinning. It does not run its own datacenters and buy racks of new drives. Secondly, it employs distributed Reed-Solomon erasure coding instead of replicated drive copies to safeguard data and this requires fewer disk drive copies to safeguard data to the same extent.
Summarizing, the paper says a corporate datacenter generates 523kg of CO2 by storing 1TB of data for three years. Hyperscalers, being more efficient, emit 251kg, but Storj produces just 12kg. Taking everything in the research paper into account, Storj can make up to 66 percent carbon savings relative to hyperscalers and 83 percent compared to corporate datacenters.