Uptime and SLA calculator
An uptime percentage is a promise about downtime written in a form that hides how much downtime it permits. This converts between the two, in both directions. It runs entirely in your browser - Nothing you type is sent anywhere.
Work backwards
If a service was down for this long in a month, what uptime does that give it?
What the nines actually buy you
| Uptime | Per month | Per year | What that feels like |
|---|---|---|---|
| 99% | 7h 18m | 3d 15h | A bad afternoon every month |
| 99.5% | 3h 39m | 1d 19h | Noticeable, and complained about |
| 99.9% | 43m 50s | 8h 46m | One incident a month, survivable |
| 99.95% | 21m 55s | 4h 23m | Most people never notice |
| 99.99% | 4m 23s | 52m 36s | Only monitoring sees it |
| 99.999% | 26s | 5m 15s | Expensive, and rarely honest |
Monthly figures use a 30-day month; yearly figures use 365 days.
Reading an SLA honestly
Once you have the minutes, the downtime cost calculator turns them into a figure somebody with a budget will respond to. The gap between 99.9% and 99.99% sounds like a rounding error and is a factor of ten. That is the entire reason uptime is quoted as a percentage rather than in minutes: 43 minutes a month is a number an operations team has to defend, and "three nines" is a number that sounds like success. Converting back is the first thing worth doing with any figure in a contract.
Check the measurement window
Almost every SLA measures monthly, and the difference matters more than the percentage does. A provider that is down for eight hours in one month has badly breached a monthly 99.9% target, and can still finish the calendar year above 99.9% overall. Any annual figure in marketing copy is quietly averaging away the month that hurt you.
Check what is excluded
Announced maintenance is excluded from nearly every commercial SLA. So, frequently, are outages attributed to a third-party network, a customer configuration error, or anything the provider designates force majeure. A service can report 100% availability for a month in which you could not use it for six hours, without anybody lying. This is also the main reason an independent measurement and a provider’s own status page disagree: Our figures exclude nothing, because if it did not answer, it was down.
Check what the remedy is worth
The remedy for a missed SLA is almost always a service credit - A percentage of that month’s fee, which you usually have to claim yourself, within a window, in writing. On a $200-a-month service, a breach that costs you a day of trading might return $20. Service credits are a governance mechanism, not compensation, and reading them as insurance is a common and expensive mistake.
Common questions
What does 99.9% uptime actually allow?
Just under 44 minutes of downtime a month, or 8 hours 46 minutes a year. It sounds close to perfect and permits an outage most people would notice. 99.99% brings that down to 4 minutes 23 seconds a month, and 99.999% to 26 seconds.
Is uptime measured monthly or yearly?
Almost every SLA measures it monthly, and that matters: A provider can be down for eight hours in March, breach its monthly target once, and still finish the year above 99.9%. Always check which window a figure refers to before comparing two of them.
Does scheduled maintenance count against uptime?
Usually not. Most SLAs exclude announced maintenance windows entirely, which is why a provider can report 100% uptime in a month you could not use the service. Our own figures make no such exclusion - if it did not answer, it was down.
What happens if a provider misses its SLA?
Typically you get a service credit, worth a percentage of that month’s fee, and you usually have to claim it yourself within a set window. The credit is almost never related to what the outage actually cost you.