Google Committed Use Discounts is a way to reduce your compute rate by committing to a set amount of servers of a certain type and in a cetain region for a period of time (1 year or 3 year). You will pay this cost whether you use the servers or not, but at a significantly discounted rate compared to the on-demand server rate. This is the equivalent of reserved instances or compute savings plan in AWS.
We can share CUD across projects, but the dimensions CUD are split across are listed below.
CUD Analysis should assume the other commitments do not end. CUD renewals should be looked at in a separate analysis so there is no confusion of CUD that cover new and existing covered usage.
Include and ping any engineering manager who will be significantly impacted by the change so they can confirm they do not expect major changes in their usage for the term of the commit. Include the spreadsheet from step 1 in the issue.
The template should include the commitment details, important high level financial details, and the engineering details about which services are most affected by the commitment.
Before the commitment is considered, the infrastructure analyst should talk with the teams that use the majority of the usage that is being committed to make sure there aren't any major changes expected during the term of the commitment. Those teams should be cc'ed in the issue and if they have any concerns voice them at that time.
Example Issue: https://gitlab.com/gitlab-com/finance/-/issues/2377
The engineering manager assigned in the issue should make the actual purchase in the billing console. If needed, the requester should join a call with them so they can both confirm the purchase is correct.
Reservations for specific node type are configured in our gitlab-com-infrastructure terraform repository. This must be updated to ensure efficient utilization of our CUD's, along with the instance choice being used in our infrastructure.
If a team is planning on making a major change to their infrastructure that would affect the commit during the term, they should check with the infrastructure analyst to assess the impact first.
