PUE data for cloud regions as an Impact Framework Dataset

[adrianco]

PUE is needed because it multiplies the energy used by a workload to account for cooling and distribution losses that are datacenter specific. Cloud providers own and operate many of their own datacenters, but also use colocation providers for smaller regions. Cloud providers may also operate datacenters for internal workloads that do not correspond directly with public cloud regions.

Google publishes PUE on a quarterly basis for its own datacenter campuses, but without providing a mapping to cloud regions. In addition many Google cloud regions are hosted in colo's that Google does not have permission to publicly disclose the PUE for.

Microsoft also publishes PUE for datacenters that it owns.

AWS has disclosed a range of PUE numbers but has no detailed information.

The intent of this work is to gather all available PUE data into a single tabular data source that can be shared by this project, then to make requests to cloud providers for enhancements that would eventually bring them all into alignment.

[adrianco]

Microsoft data is here https://datacenters.microsoft.com/globe/fact-sheets

[adrianco]

Google data is here https://www.google.com/about/datacenters/efficiency/

[adrianco]

AWS discussion here https://perspectives.mvdirona.com/2009/06/pue-and-total-power-usage-efficiency-tpue/
Range provided here https://aws.amazon.com/blogs/industries/four-trends-driving-global-utility-digitization/

[adrianco]

Data for Azure, Google and AWS has been extracted into a google sheet and it is proposed that this be exported via the GSF Impact Framework https://github.com/Green-Software-Foundation/if

[adrianco]

Google sheet for the current data set is here https://docs.google.com/spreadsheets/d/1-bui2VIBvunmyO4JLvB3BxponGaV0BctBlfAPvOg0SQ/edit?usp=sharing

[adrianco]

PUE data has been merged into the Carbon Impact Framework dataset since it's indexed by the same inputs - date, provider, cloud region.

[adrianco]

Closing this issue, new issues have been opened against each of the three main cloud providers to request bringing their data into alignment.
Azure #27
GCP #28
AWS #29