Q: How do I actually calculate the downtime budget my SLA allows?

Q: How do I actually calculate the downtime budget my SLA allows?

A: Memorize these per-month figures, they're the whole game.

— 99.9% ("three nines") = 43m 49s of downtime per month.
— 99.95% = 21m 54s.
— 99.99% ("four nines") = 4m 22s.
— 99.999% = 26 seconds. Basically unachievable with manual response.

Now the playbook: take your promised number, find the monthly budget above, then subtract planned maintenance. What's left is your real margin for surprises. If you promise 99.99% but your detect-plus-respond time is 10 minutes, one incident blows the whole month.

Follow-up everyone misses: define whether your SLA counts partial degradation as downtime. "Slow checkout" vs "checkout 500ing" need a written rule before an incident, not during one.

Got a question? Drop it in the comments.