A collaborative collection of information required to provide estimates of carbon usage for any cloud services, taking into account both initial setup and ongoing running costs.

Goals:

• enable developers to understand the environmental impact of their infrastructure
• discover gaps in the environmental information published by providers

What we found:

• Information available at a marketing level, backed by research but not easy to use to drive decisions, can be contradictory.
• Hard to evaluate at an individual or customer level.
• Lack of clarity around carbon offsetting.

Questions:

• How is environmental impact measured?
• At what granularity is information provided?
• How easy is this information to access?

Independent and academic research

//TODO

Amazon

Metrics Used

Power Usage Effectiveness

This is the total power that is delivered to the server, storage, and networking in a data center (this is called critical power), divided by the total power that is brought to a data center (this is called total power).

Carbon Intensity

Grams of carbon emissions per kWh of energy used The best estimate of Amazon's CI is from June 2015 where the average power mix carbon intensity is 393 grams/kWh

Granularity

You would have to calculate the kWh of each service based on the CPUs used in the services in order to compare the services

Ease of Access

Not all of the services give an way to calculate the usage. For those that do, it's still not simple to calculate.

Microsoft

(MS Sustainability homepage)

A lot of focus on Microsoft's impact at a corporate level, including published reports such as 2018 GRI Index report, CDP Climate change response 2018

A strong focus on the benefits of cloud over client-site solutions, possibly backed by data here. Could these measurements be used to compare cloud solutions?

Information doesn't seem to be available at the granularity of individual customers or developers?

Google

//TODO Carbon offsetting, buying renewable energy (are prices equivalent?), waste commitments - diverted from landfill, to where?

IBM

Other avenues to explore

• Shared networking infrastructure
• Individual overheads, e.g. dev machines