Rightsizing and Autoscaling Economics

1. Paying for Peak, Using the Average

Provisioned capacity is paid for by the hour whether or not anything uses it. Demand is not flat. The gap between those two facts is where most compute spend goes.

The number that measures the gap is utilization: the fraction of what you are paying for that is actually doing work. Across most organizations it is far lower than the people paying for it expect, and the reason is rarely carelessness.

• Capacity is sized for peak, because that is the hour that has to work.
• Headroom is added for a failed instance and for a deploy.