Capacity planner
Work out how many instances you need for peak traffic, with room for growth and a failure.
How it works
Peak load is average traffic, grown, times the peak factor. Each instance runs at your target utilisation, and spare instances cover a failure or a deploy.
instances = ⌈peak ÷ (capacity × target)⌉ + spareQuestions
Why not run at 100%?
Latency climbs sharply as utilisation nears 100%. Most services aim for 50 to 70%.
How many spare instances?
At least one, so losing an instance or rolling a deploy doesn’t overload the rest.