Hi Olivier,
Thanks for your question! You are pretty much spot-on. We parse the usage type from the CUR report where we can determine the machine type based on the S3 configuration. Depending on the usage type, we can determine whether the underlying storage is using an SSD or HDD so that we can use the appropriate coefficient when estimating. You can check out a list of those services in
this file as an example of how we go about this for all service that use associated storage, including S3. From there, we estimate the energy usage based on the allocated storage and multiply it by the replication factor for that appropriate service.
We realized that some of the specifics may be missing from the
Storage Methodology, which is why we always welcome feedback. If you feel this documentation could be improved to elaborate on this, we would welcome
opening a PR to do so if you're still interested.
Thanks,
The CCF Team at Thoughtworks