Comparison
HeadroomvsOverprovisioning
Headroom
you run at sixty percent so a zone failure or a traffic spike does not immediately become an outage.
Deliberate unused capacity held to absorb failures, spikes and the time it takes to add more. The right amount is set by what you must survive without scaling — commonly the loss of one zone — plus however long provisioning actually takes. It is the least glamorous line on a cloud bill and the first thing cut in a cost exercise, usually by someone who has not connected it to the last incident.
Full entry →Overprovisioning
every service asked for four cores because that is what the template said, and the fleet averages eight percent CPU.
Running more capacity than the workload needs, usually through copied defaults, fear of the last incident, or requests set from peak rather than from measurement. It is the largest single source of waste in most estates and the easiest to quantify, since utilisation data already exists. It is distinct from deliberate headroom, and telling the two apart is the first task of any cost exercise.
Full entry →