Uptime SLA calculator
An SLA percentage is an error budget written in a way that hides how small it is. Enter a target below to see the downtime it actually permits — or enter the minutes you lost to see the uptime you actually delivered.
| Period | Allowed downtime |
|---|---|
| Per day | 1m 26s |
| Per week | 10m 5s |
| Per month | 43m 50s |
| Per quarter | 2h 11m 29s |
| Per year | 8h 45m 58s |
That is 99.9019% uptime for the month.
How the numbers are calculated
Allowed downtime is simply the complement of the target multiplied by the length of the period: at 99.9%, 0.1% of a 30.44-day month is 43 minutes 12 seconds. Months here use the average calendar month (365.25 ÷ 12 days) rather than 30 days, because SLAs are measured against real calendar months and the difference is material at four nines.
Frequently asked questions
- How much downtime does 99.9% uptime allow?
- 99.9% uptime ("three nines") allows about 43 minutes 50 seconds of downtime per average calendar month, or roughly 8 hours 46 minutes per year. If your SLA measures a flat 30-day month instead, the budget is 43 minutes 12 seconds. Either way, a single bad deploy can spend all of it in one go.
- What is the difference between 99.9% and 99.99%?
- One order of magnitude. 99.9% gives you about 43 minutes of downtime a month; 99.99% gives you about 4 minutes 23 seconds. Going from three nines to four nines usually means removing every manual step from recovery, because a human cannot even read the alert in four minutes.
- Is uptime measured per month or per year?
- Read the SLA. Most providers measure per calendar month, which resets the budget every month and caps the credit you can claim. An annual measurement is more forgiving to the provider: one long outage can be averaged away across twelve months.
- Does a 1-minute check interval change my measured uptime?
- It changes how much of it you can see. With 5-minute checks an outage shorter than five minutes can fall entirely between two probes and never appear in your records. The outage still happened — your monitoring just missed it.