How much downtime is 99.9% uptime
99.9% uptime allows 8 hours 45 minutes 57.6 seconds of downtime per year, 43 minutes 49.8 seconds per average month, 10 minutes 4.8 seconds per week and 1 minute 26.4 seconds per day.
The 0.1% is the whole story. A year of 365.25 days is 525,960 minutes, and 0.1% of that is 525.96 minutes. That is your downtime budget for the year, and every number below is the same budget cut into smaller periods.
99.9 uptime means how much downtime
- Per year: 525.96 minutes, which is 8.766 hours or 8 h 45 min 57.6 s.
- Per average month of 30.4375 days: 43.83 minutes, or 43 min 49.8 s.
- Per week: 10.08 minutes, or 10 min 4.8 s.
- Per day: 86.4 seconds, or 1 min 26.4 s.
99.9 uptime hours per year
8.77 hours, if the year is 365.25 days. Use 365 days and it drops to 8.76 hours; use 366 and it is 8.78. The difference is about a minute at most, so any of them is fine for planning. What matters is that 8.77 hours is one working day. One bad migration that takes a full day to roll back spends the whole year.
99.9 uptime hours per month
0.73 hours, or 43.8 minutes, for an average month. Real months are not average, and contracts measure calendar months, so the budget moves: 40.32 minutes in a 28-day February, 43.2 minutes in a 30-day month and 44.64 minutes in a 31-day month. February gives you 4.3 minutes less than March for the same promise.
The math, for any target
awk -v target=99.9 'BEGIN {
down = (100 - target) / 100
year = 365.25 * 24 * 60 # minutes in a year
printf "per year: %.1f min (%.2f h)\n", year * down, year * down / 60
printf "per month: %.1f min\n", year / 12 * down
printf "per week: %.1f min\n", 7 * 24 * 60 * down
printf "per day: %.1f s\n", 24 * 60 * 60 * down
}'
# per year: 526.0 min (8.77 h)
# per month: 43.8 min
# per week: 10.1 min
# per day: 86.4 s- 99%: 3 d 15 h 39.6 min a year, 7 h 18.3 min a month.
- 99.5%: 1 d 19 h 49.8 min a year, 3 h 39.2 min a month.
- 99.95%: 4 h 23 min a year, 21.9 min a month.
- 99.99%: 52.6 min a year, 4.4 min a month.
What 99.9% looks like to a 5-minute check
A monitor that checks every 5 minutes makes 8,766 checks in an average month. 0.1% of that is 8.77, so 8 failed checks leave you at 99.909% and the ninth puts you at 99.897%. Logdash computes uptime the same way, successful checks over all checks, so the number is only as precise as the interval. An outage shorter than 5 minutes can fall between two checks and cost you nothing on paper. At one check a minute the month is 43,830 checks and 43 failures fit in the budget.
99.9 uptime SLA
A 99.9% SLA from a provider covers their service, measured their way, and pays back in credits on their bill. It says nothing about the app you built on top. If a request needs three services and each one hits exactly 99.9%, the most you can promise is 0.999 cubed, 99.7%, which is about 2 hours 11 minutes a month. Your own code and deploys come out of what is left.
- Add the URL Create a service in Logdash and paste the address of your health endpoint, or your homepage if the site is static. The first check runs straight away, so a wrong path shows up in seconds rather than during an incident.
- Pick the interval At 99.9% the monthly budget is 43.8 minutes, and a 5-minute interval can spend 5 of those, 11%, before the first failed check. Every minute on Builder brings that to 1 minute. Every 15 seconds on Pro brings it to 15 seconds.
- Break it on purpose Connect a Telegram channel, then stop the app or make the endpoint return 503. An alert you never tested is an alert you cannot trust. On the next check the monitor flips to down and a Telegram alert lands on your phone with the monitor name, the status code and the error.