99.9% uptime is commonly called "three nines".
Allowed downtime at 99.9% uptime
Promising an SLA you can't see?
A calculator tells you the budget. ObserveOne tells you when you're about to blow it: uptime monitoring, incident alerts, and public status pages from one dashboard.
Picking a monitoring tool?
Side-by-side breakdowns, no fluff.
Common comparisons
Alternatives worth a look
Common SLA uptime tiers
| Uptime | Nines | Downtime / month | Downtime / year |
|---|---|---|---|
| 90% | one nine | 3d 1h 3m 0s | 36d 12h 36m 0s |
| 95% | - | 1d 12h 31m 30s | 18d 6h 18m 0s |
| 99% | two nines | 7h 18m 18s | 3d 15h 39m 36s |
| 99.5% | - | 3h 39m 9s | 1d 19h 49m 48s |
| 99.9% | three nines | 43m 50s | 8h 45m 58s |
| 99.95% | three and a half nines | 21m 55s | 4h 22m 59s |
| 99.99% | four nines | 4m 23s | 52m 36s |
| 99.999% | five nines | 26.3s | 5m 16s |
Frequently Asked Questions
What does a 99.9% SLA actually mean?
It means the service is allowed to be down for 0.1% of the agreed reporting period. For a calendar month that works out to about 43 minutes; for a full year it is roughly 8 hours and 46 minutes. The exact figure depends on whether the SLA is measured monthly, quarterly, or annually.
Why do vendors talk about "nines"?
It is shorthand for how many 9s appear in the uptime percentage: 99% is "two nines," 99.9% is "three nines," 99.99% is "four nines," and so on. Each additional nine cuts allowed downtime by roughly 10x, which is why going from three nines to four nines is a much bigger operational commitment than the percentages alone suggest.
Does this calculator account for scheduled maintenance?
No. It assumes the SLA covers continuous 24/7 uptime, which is the most common way SLAs are written. If your contract excludes a maintenance window, subtract that window from the period before comparing it against the allowed-downtime figures here.
How is the monthly and yearly figure calculated?
Using a calendar-average month (365.25 days ÷ 12) and a 365.25-day year, so leap years are accounted for rather than assuming every month is exactly 30 days. This is the same convention used by most public SLA calculators.
What SLA should I actually promise customers?
Promise what your infrastructure and on-call process can realistically hold, not what sounds impressive. Most SaaS products land between 99.9% and 99.95%; anything above 99.99% usually requires multi-region failover and a mature incident-response setup. Committing to more nines than you can operationally support just turns every incident into a breach.