99.99% uptime (four nines)
99.99% uptime allows 4m 19s of downtime in a 30-day month and 52m 34s in a year. Per day that is 8.6s; per week, 1m. Here is what fits in that budget, who promises 99.99%, and how to measure whether you're meeting it.
How much downtime is 99.99%?
Uptime is the share of time a service answers. The remaining 0.01% is the downtime budget, and it only looks small until it's written in minutes. The month figure uses 30 days; the year uses 365.
| Period | Allowed downtime |
|---|---|
| Per day | 8.6s |
| Per week | 1m |
| Per month (30 days) | 4m 19s |
| Per year | 52m 34s |
What 99.99% means in practice
Four minutes a month is less than the time it takes a person to read an alert, open a laptop and restart a service. Four nines is only possible when failures are handled by machines: a load balancer that drops a bad server, a database that fails over by itself, deploys that roll out gradually and roll back on error.
Monitoring still matters at this level, but its job changes: it's there to tell you the automatic failover didn't work, and to produce the number you report.
Who promises 99.99%
99.99% is promised by major cloud providers for multi-zone or multi-region deployments of their core services (object storage, load balancers, DNS), and by payment and infrastructure companies whose customers' revenue depends on them. It's typically backed by service credits.
Few small SaaS products promise it, and those that do usually run on providers who promise it to them.
How to measure it
Uptime is only a number if something is measuring it from outside, often enough to see short outages. heyUptime checks every 30 seconds (60 on Starter), confirms a failure with a second check before counting it, keeps 90 days of history on the status page, and sends a weekly summary with the uptime percentage per monitor. That's the figure to compare with 99.99%.
To turn any percentage into minutes, or minutes into a percentage, use the uptime calculator.
- Checks every 30 to 60 seconds; shorter outages than that can't be seen by any external monitor.
- Confirmed failures: a single slow response doesn't count as downtime.
- Maintenance windows are recorded but not counted as outages on the status page.
- The weekly summary reports uptime and response time per monitor for the week.
Is 99.99% enough for you?
In short: 99.99% is a target for infrastructure that other people's businesses depend on. For a website or a typical SaaS product, 99.9% to 99.95% is the realistic promise.
Measuring four nines from outside needs checks at least every 30 seconds and a definition of "down" that excludes single slow responses; a 90-day history and a weekly summary give the trend. Be aware that external monitoring measures what visitors see, including DNS and the network between, which is the honest number but not always the one in the provider's SLA.
Questions
How much downtime is 99.99% per month?
4 minutes and 19 seconds in a 30-day month; 1 minute a week; 8.6 seconds a day.
How much downtime is 99.99% per year?
52 minutes and 34 seconds.
Can a single server reach 99.99%?
Rarely, and not on purpose. One security reboot or one crashed process uses most of a month's 4m 19s. Four nines needs more than one of everything and automatic failover.
What does "four nines" mean?
99.99% uptime: four nines in the number. Three nines is 99.9% (43 minutes a month), five nines is 99.999% (26 seconds a month).
See it on your own site.
heyUptime checks every 30 seconds, confirms before alerting, and tells you why it's down in plain words. Free for 14 days, no card.
Related
Updated October 8, 2026.