SLA Uptime & Downtime Calculator
Turn an uptime percentage into real downtime, plus why 99.9% means 43 minutes a month or 8.8 hours a year depending on nothing but the window.
Months are taken as 30 days and years as 365, so a real contract with calendar months will differ slightly. Nothing is uploaded.
A percentage without a window says very little
99.9% is 43m 12s over a month and 8h 45m 36s over a year. Same number, same service, twelve times the allowance — so an agreement that does not say which window it means has not really committed to anything.
| Target | Per day | Per week | Per month | Per quarter | Per year |
|---|---|---|---|---|---|
| 90% | 2h 23m 60s | 16h 47m 60s | 2d 23h 59m | 9d 2h 23m | 36d 11h 59m |
| 95% | 1h 12m | 8h 24m | 1d 12h | 4d 13h 12m | 18d 6h |
| 99% | 14m 24s | 1h 40m 48s | 7h 12m | 21h 50m 24s | 3d 15h 36m |
| 99.5% | 7m 12s | 50m 24s | 3h 36m | 10h 55m 12s | 1d 19h 48m |
| 99.9% | 1m 26s | 10m 5s | 43m 12s | 2h 11m 2s | 8h 45m 36s |
| 99.95% | 43s | 5m 2s | 21m 36s | 1h 5m 31s | 4h 22m 48s |
| 99.99% | 9s | 1m | 4m 19s | 13m 6s | 52m 34s |
| 99.999% | 864 ms | 6s | 26s | 1m 19s | 5m 15s |
One outage, two verdicts
A single 4h outage comfortably clears 99.9% measured over a year, and blows the same target measured over a month by a factor of 5.6. Nothing about the incident changed. That is the practical reason the window belongs in the sentence: the identical failure is either within tolerance or five times over it, depending on a word that often goes unsaid.
Each extra nine costs a factor of ten
Every nine divides the remaining budget by ten, exactly. So the step from three nines to four is not an incremental tightening — it removes nine tenths of the room you had left.
| Target | Per year | Against the previous row |
|---|---|---|
| 99% | 3d 15h 36m | — |
| 99.9% | 8h 45m 36s | 10× tighter |
| 99.99% | 52m 34s | 10× tighter |
| 99.999% | 5m 15s | 10× tighter |
Which is why the nines are counted rather than the percentage read. Going from 99% to 99.9% and from 99.9% to 99.99% look like small moves on paper — nine tenths of a per cent, then nine hundredths — and each is the same tenfold jump in what the engineering has to achieve.
A chain is worse than any of its links
Availabilities multiply when every component has to work. Five services at 99.9% each deliver 99.501% together — very nearly five times the downtime. To reach 99.9% end to end across five, each has to hold 99.980%.
| Components in series | All at 99.9%, you get | For 99.9% overall, each needs |
|---|---|---|
| 1 | 99.900% | 99.9000% |
| 2 | 99.800% | 99.9500% |
| 3 | 99.700% | 99.9667% |
| 5 | 99.501% | 99.9800% |
| 10 | 99.004% | 99.9900% |
| 20 | 98.019% | 99.9950% |
The requirement rises with every link, and quickly stops being reasonable: twenty components in series each need better than 99.995% for the whole to manage three nines. Which is an argument about architecture rather than about operations — the number of things that must all work is the number that decides what any of them has to promise.
How to use
- Set an uptime target.
- Choose the window it is measured over.
- Enter an outage length to test against it.
- Every window is shown so you can see the difference.
Frequently asked questions
How much downtime does 99.9% allow?
43 minutes and 12 seconds over a month, or 8 hours 45 minutes over a year. Same number, same service, twelve times the allowance — which is why an agreement that does not say which window it means has not committed to very much.
Does the measurement window really matter that much?
It decides the verdict. A single four-hour outage comfortably clears 99.9% measured annually and overruns the same target monthly by a factor of 5.6. Nothing about the incident changes; only the sentence it is measured against.
What does one more nine cost?
Nine tenths of whatever budget is left. 99% allows three and a half days a year, 99.9% allows nine hours, 99.99% allows 53 minutes and 99.999% allows five. Each step looks small written as a percentage and is a tenfold tightening of the engineering.
Why is my system less available than its components?
Because availabilities multiply when everything has to work. Five services at 99.9% each deliver 99.50% together, which is very nearly five times the downtime of any one of them. Ten deliver 99.00%.
So what does each component need?
More than the target it serves. For five components in series to reach 99.9% end to end, each must hold 99.98%. For twenty, each needs better than 99.995%. That is an argument about architecture rather than operations — the number of things that must all work decides what any of them has to promise.
Is 100% uptime a target?
Not a meaningful one. It has no number of nines, allows no downtime at all in any window, and leaves no room for the deployments and maintenance that keep a system working. The useful question is which window and how many nines within it.
How are months and years counted here?
A month as 30 days and a year as 365. A real contract with calendar months will differ slightly — February at 99.9% allows about four minutes less than a 30-day month — so treat these as the shape of the answer rather than the last word on a specific agreement.
Does this send anything anywhere?
No. Everything is computed in your browser, and nothing is uploaded.
🔒 This tool runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.