Has anyone been successful in implementing a FinOps Cost Optimization policy to turn down idle cloud instances during off hours (nights/weekends)? DId you start with non-prod environments? How are you handling logging/monitoring alerts during that time?
Sort by:
No, not yet. We are looking for a suitable FinOps tool to make cloud and on-premise resource consumption comparable.
"Successful" is a funny word. We were able to use Harness to demonstrate that it was possible to use their Stop/Start technology to turn off not-in-use VMs, but the process to get to that point was so laborious that we declined to move forward with a full implementation. And yes, we started in non-Prod.
nops.io was amazing when I used them. Let you automate so much
our team has been wildly successful in reducing our cloud spend by setting up schedules for off-hours. We started with our largest non-production environments and continue to chip away at our cloud based application environment to find savings. Logs and monitoring will need to change as you deploy your schedules and that needs coordination with both infrastructure and security teams.
We pursued this optimization couple of years ago for our VMs on Azure. However, considering most of our production load is on reserved instances, there is no prod centric cost benefit. For non-prod, we had to work on a more thought-out schedule of VM types (based on our learnings) as some non-prod do not lend themselves well to such ramp downs. E.g., UAT/QA instances which are being used by global business teams for testing. They tend to be used across time zones for us leaving a very small window of idle time. Dev boxes are a better bet for such optimization.