Downtime for common SLAs
| SLA | Day | Week | 30-day month | 365-day year |
|---|---|---|---|---|
| 90% | 2h 24m | 16h 48m | 3d | 36d 12h |
| 95% | 1h 12m | 8h 24m | 1d 12h | 18d 6h |
| 99% | 14m 24s | 1h 40m 48s | 7h 12m | 3d 15h 36m |
| 99.5% | 7m 12s | 50m 24s | 3h 36m | 1d 19h 48m |
| 99.9% | 1m 26s | 10m 5s | 43m 12s | 8h 45m 36s |
| 99.95% | 43.2s | 5m 2s | 21m 36s | 4h 22m 48s |
| 99.99% | 8.64s | 1m | 4m 19s | 52m 34s |
| 99.999% | 0.86s | 6.05s | 25.9s | 5m 15s |
How to work out allowed downtime
Take the SLA away from 100% and multiply by the length of the window. At 99.9%, 0.1% of a 30-day month (2,592,000 seconds) is 2,592 seconds, or 43 minutes 12 seconds.
- 99.9% of a 30-day month: 43m 12s
- 99.9% of a 31-day month: 44m 38s
- 99.9% of an average month (365 ÷ 12 = 30.42 days): 43m 48s
- 99.9% of a 365-day year: 8h 45m 36s
- Two services that both have to be up, each at 99.9%: 99.9% × 99.9% = 99.8001%
These match the availability table in Google's SRE book, which uses a 30-day month, a 90-day quarter and a 365-day year.
Which window to write into an SLA
The same outage can pass or fail depending on the window. A monthly SLA gives each month its own allowance, so one bad night can breach it. A yearly SLA lets a bad month be made up later, but you may not know you are in trouble until late in the year. Calendar months also differ: February has 28 or 29 days and the longest months 31, so a fixed "43 minutes a month" and "99.9% each month" are not quite the same promise.
Planned maintenance
Many SLAs leave out maintenance that was announced in advance. When you tick that option, the calculator takes the maintenance minutes off the period before working out the budget, so a 4-hour window in a 30-day month leaves a 99.9% budget of 42m 58s instead of 43m 12s.
Questions
How much downtime is 99.9% uptime?
About 1 minute 26 seconds a day, 10 minutes 5 seconds a week, 43 minutes 12 seconds in a 30-day month, 44 minutes 38 seconds in a 31-day month and 8 hours 45 minutes 36 seconds in a 365-day year.
Why do sites give different numbers for a month?
Because they use different month lengths. A 30-day month allows 43m 12s at 99.9%, a 31-day month 44m 38s, and an average month (365 ÷ 12 days) 43m 48s. Say in the contract which one you mean, or measure each calendar month as it comes.
Does a monthly SLA mean the same as a yearly one?
No. 99.9% a year allows 8h 45m 36s in total, so one 3-hour outage keeps you inside it. Measured monthly, the same outage breaks that month's 43m 12s allowance even if the rest of the year is perfect.
How do I work out the uptime of services that depend on each other?
If every service has to be up for yours to work, multiply their availabilities: 99.9% × 99.9% = 99.8001%. This assumes they fail independently. A chain is never more available than its weakest part.
What is an error budget?
The downtime your SLA allows in the current window, less what you have used. In Google's SRE book, new releases go out while budget remains; when it is used up, releases pause while the team works on reliability.
Is 100% uptime possible?
The calculator will show 0 seconds allowed, but Google's SRE book says 100% is probably never the right reliability target: it is impossible to achieve, and more than users want or notice.
Does anything I type leave my browser?
No. The sums run in your browser and your inputs are saved only in this browser. A share link keeps everything after the # in the address, and browsers never send that part to a server. "Clear my data" removes what is saved here.
Sources
Last checked 4 October 2026. The figures are arithmetic; these sources define the terms and give a published table to check against.
- Google SRE book, Appendix A: availability table
- Google SRE book, chapter 3: Embracing risk (error budgets)
Changes: 4 October 2026: first published.