Skip to main content
HostTracker
HOST-TRACKER
website monitoring service
Status pages Instant check Price
Login Register

Uptime and SLA calculator

Convert an SLA uptime percentage into the downtime it actually allows per day, week, month and year, or work backwards from the downtime you observed to the percentage it produced. It all runs in your browser: nothing is sent anywhere and no account is needed.

What a percentage really costs you

A 99.9% target sounds close to perfect until you convert it: it still allows about 44 minutes of downtime every month and nearly nine hours a year. Reading the percentage as time is the fastest way to see whether a service level matches what your users will actually tolerate.

Find out more

How the periods are counted

A day and a week are exact. A month is counted as 30.44 days and a year as 365.25 days, the long-run averages that make the yearly figure agree with every published SLA table. Always convert against the period your contract is written for, because the same percentage buys very different amounts of time.

Find out more

An estimate is not a measurement

This tool converts a number you typed. It cannot tell you what your uptime actually was. That figure comes from checks running against your service around the clock, from more than one network, with every failure confirmed from a second location before it counts as downtime.

Find out more

What 99.9%, 99.95% and 99.99% uptime allow

Uptime target Downtime per month Downtime per year
99.9% 43m 50s 8h 45m 57s
99.95% 21m 55s 4h 22m 58s
99.99% 4m 23s 52m 35s
99.999% 26s 5m 15s

Based on a 30.44-day average month and a 365.25-day average year, the same long-run averages the calculator above uses.

A percentage in a contract only holds up as an SLA when it is backed by monitoring that runs independently of your own infrastructure - see how distributed availability monitoring confirms it.

These figures are an estimate, not a measurement: they are arithmetic on a percentage you typed, while your real uptime is whatever independent checks recorded over the period. How uptime is actually calculated or start measuring yours from 300+ locations.

Key takeaways

  • This calculator converts an uptime percentage into the downtime it allows per day, week, month and year, and converts a downtime figure back into a percentage.
  • It is arithmetic in your browser: it runs no check, contacts no server and needs nothing about your site. A month is an average 30.44 days and a year 365.25 days.
  • On that basis 99.9 percent uptime allows about 43 minutes 50 seconds of downtime a month, and 99.99 percent allows about 4 minutes 23 seconds.
  • It is free, needs no login and no credit card.
  • For continuous checks, see the distributed availability monitoring page: measured uptime comes from a monitor's own recorded history, not from a target figure.

Frequently asked questions.

Subtract the downtime from the length of the period, divide by the length of the period and multiply by 100. For example, 0.1% of an average year is 31,557.6 seconds (8h 45m 57.6s), so a year with that much downtime has 99.9% uptime. The second panel of the calculator does this for you: enter the downtime, choose the unit and the period, and it shows the percentage. It is arithmetic on figures you provide, not a measurement.
A 99.9% target allows 1m 26s of downtime a day, 10m 4s a week, 43m 50s in an average month of 30.44 days and 8h 45m 57s a year. Counted against calendar months, it is 40m 19s in a 28-day February, 43m 12s in a 30-day month and 44m 38s in a 31-day month.
A 99.99% target allows 8.64 seconds of downtime a day, about one minute a week, 4m 23s in an average month and about 52.6 minutes (52m 35s) a year. A single five-minute incident therefore uses more than the whole month's 4m 23s allowance.
It means 99.999% uptime, which allows under one second of downtime a day (0.864 seconds), 26 seconds in an average month and 5m 15s a year. It is usually reserved for systems with redundant components and automatic failover, because a human response to an alert takes longer than the whole monthly budget.
It is a low bar for a public website. 99% allows 7h 18m 20s of downtime in an average month, close to a full working day, and 3 days 15 hours 39 minutes a year. 99.9% is a more common baseline for commercial services.
0.1% of an average year is 31,557.6 seconds, which is 8h 45m 57.6s. The calculator cuts durations off at the whole second and never rounds up, so it never shows more downtime than a target allows. Rounded to the nearest minute, that same figure becomes 8h 46m.
The calculator treats every second of downtime alike, but contracts often do not. These points are common contract practice, not legal rules; your own agreement decides.
  • Announced maintenance is usually excluded; unannounced maintenance usually counts.
  • Short blips may be ignored until an outage lasts a minimum number of consecutive minutes.
  • Partial outages are handled in very different ways, from complete unavailability only to a share of users or requests affected.
  • Causes outside the provider's control, such as your own code or attacks, are commonly excluded.
  • Responses slower than a stated threshold can count as downtime.
To measure an SLA from outside the provider's network, see SLA monitoring.
If your service needs several others up at the same time, its best possible availability is the product of theirs, assuming they fail independently. With web hosting at 99.95%, a managed database at 99.99% and a payment API at 99.9%: 0.9995 x 0.9999 x 0.999 = 0.99840064995, or 99.8401% rounded. That allows 1h 10m 6s of downtime a month, almost exactly the three allowances (21m 55s, 4m 23s and 43m 50s) added together. If each provider used its full allowance in separate incidents, you could not promise your own users even 99.9%.
Yes. Redundancy works the opposite way from a chain of dependencies: two independent servers that can each carry all the traffic, each at 99.9%, are both down only 0.1% x 0.1% = 0.0001% of the time, which is 99.9999% for the pair. That holds only if failover is automatic, and whatever switches between them (a load balancer, a DNS failover) becomes one more component in the chain.
A monitor cannot time an outage more precisely than its check interval: it learns of downtime at the first failed check and of recovery at the first successful one. Each incident's measured length can be off by up to one interval either way, and an outage shorter than the interval can fall between two checks and never be recorded. At 5 minutes that error is 114.1% of a 99.99% month's allowance (4m 23s), so that interval cannot tell you whether the target was met. Proving a target needs an interval well below its monthly allowance, and confirmation from more than one location.
Yes, any calculation can be shared or bookmarked. The page address updates as you type, so copying it keeps your numbers. Adding ?uptime=99.95 opens the calculator at 99.95%, and ?downtime=45&unit=m&period=month fills in 45 minutes over a month (unit=m means minutes, unit=h hours).

Part of HostTracker's website monitoring software.