99.9% uptime allows 43 minutes 12 seconds of downtime in an exact 30-day month and 8 hours 45 minutes 36 seconds in a 365-day year. Enter any target for an immediate conversion, or use reverse, comparison, and error-budget modes for deeper analysis.
Results are ready.
Enter downtime and the full measurement period. The result is the availability that downtime represents.
Choose at least two targets. The same fixed periods are used for every row.
| Target | Day | Week | 30 days | 90 days | 365 days |
|---|
Important: burn rate = observed failure rate ÷ sustainable failure rate. Above 1× predicts a future SLO miss if that rate is sustained; it does not by itself prove the current SLA window has already been breached. Budget consumption ratio compares failures already observed with the full compliance-period allowance.
1 day = 86,400 seconds; week = 7 days; fixed month = 28, 30, or 31 days; fixed quarter = 90 days; common year = 365 days. The average calendar month is 365.2425 ÷ 12 days. Exact dates use elapsed UTC midnights with the end date exclusive.
Calculations use unrounded floating-point seconds. Durations display to 0.01 second when useful; percentages display up to six decimal places. A leap year is only included when exact dates span February 29; the fixed year remains 365 days.
Time mode treats entered downtime as fully unavailable. Request mode uses failed eligible events and can better represent partial outages. Maintenance, degraded service, exclusions, and weighting are included only if you include them in the entered figures.
Accountable organization: Starlight Robotics. Last technically reviewed: . Method follows Google SRE's error-budget definitions, availability table, and Google Cloud's failure-rate burn-rate definition.
All monthly values below use an exact 30-day month, the quarter uses 90 days, and the year uses 365 days. The current basic-calculator target is highlighted when it matches a row.
| Target | Day | Week | 30-day month | 90-day quarter | 365-day year |
|---|---|---|---|---|---|
| 99% two nines | 14m 24s864 seconds | 1h 40m 48s6,048 seconds | 7h 12m25,920 seconds | 21h 36m77,760 seconds | 3d 15h 36m315,360 seconds |
| 99.5% | 7m 12s432 seconds | 50m 24s3,024 seconds | 3h 36m12,960 seconds | 10h 48m38,880 seconds | 1d 19h 48m157,680 seconds |
| 99.9% three nines | 1m 26.4s86.4 seconds | 10m 4.8s604.8 seconds | 43m 12s2,592 seconds | 2h 9m 36s7,776 seconds | 8h 45m 36s31,536 seconds |
| 99.95% | 43.2s43.2 seconds | 5m 2.4s302.4 seconds | 21m 36s1,296 seconds | 1h 4m 48s3,888 seconds | 4h 22m 48s15,768 seconds |
| 99.99% four nines | 8.64s8.64 seconds | 1m 0.48s60.48 seconds | 4m 19.2s259.2 seconds | 12m 57.6s777.6 seconds | 52m 33.6s3,153.6 seconds |
| 99.999% five nines | 0.864s0.864 seconds | 6.048s6.048 seconds | 25.92s25.92 seconds | 1m 17.76s77.76 seconds | 5m 15.36s315.36 seconds |
| 99.9999% six nines | 0.0864s0.0864 seconds | 0.6048s0.6048 seconds | 2.592s2.592 seconds | 7.776s7.776 seconds | 31.536s31.536 seconds |
period × (1 - target ÷ 100). At 99.9%, the unavailable fraction is 0.001, so 2,592,000 seconds × 0.001 = 2,592 seconds in 30 days.
(period - downtime) ÷ period × 100. Use the same units for both values. This measures observed availability; it does not decide which incidents a contract excludes.
A time budget is allowed downtime. A request budget is eligible events × (1 - SLO). Consumption ratio is observed bad time or events divided by that full-period allowance. True burn rate instead divides the observed lookback failure rate by the SLO's sustainable failure rate.
An SLI is the measured reliability signal, such as successful requests ÷ eligible requests. An SLO is the target for that signal. An SLA is an agreement that may attach remedies and define different measurement or exclusion rules.
Planned maintenance, customer-caused failures, third parties, force majeure, degraded service, and claim procedures vary by agreement. Apply the actual contract; this tool provides transparent arithmetic, not legal interpretation.
2,592,000 × 0.0001 = 259.2 seconds, or 4 minutes 19.2 seconds per 30 days. For 365 days: 31,536,000 × 0.0001 = 3,153.6 seconds, or 52 minutes 33.6 seconds.
In 30 days: (720 hours - 2 hours) ÷ 720 × 100 = 99.722222%. A 99.9% target would be missed because two hours exceeds its 43-minute 12-second allowance.
(720 hours - 4 hours) ÷ 720 × 100 = 99.444444%. Rounded to four decimal places, the required target is 99.4444%.
Sixty minutes in a 30-day budget of 43 minutes 12 seconds is 60 ÷ 43.2 = 138.89% consumed, exceeding the allowance by 16 minutes 48 seconds. Over a one-day lookback, the failure-rate burn rate is (60 ÷ 1,440) ÷ 0.001 = 41.67×.
Multiply the period duration by the unavailable fraction: allowed downtime = period × (1 - target / 100). A 99.9% target has a 0.1% unavailable fraction.
At 99.9%, an exact 30-day month allows 43 minutes 12 seconds. A 365-day year allows 8 hours 45 minutes 36 seconds. An average calendar month of 365.2425 days divided by 12 allows about 43 minutes 49.7 seconds.
99.99% permits one tenth as much downtime as 99.9%. Over 30 days, the allowance falls from 43 minutes 12 seconds to 4 minutes 19.2 seconds, a reduction of 38 minutes 52.8 seconds.
Three nines means 99.9%, four nines means 99.99%, and five nines means 99.999%. Each extra nine cuts the unavailable fraction, and therefore allowed downtime, by a factor of ten.
Subtract downtime from the total period, divide by the total period, then multiply by 100. Four hours of downtime in an exact 30-day period corresponds to 99.4444% uptime.
It depends on the SLA or SLO policy. Some agreements exclude approved maintenance windows; others count every unavailable minute. Apply the inclusions and exclusions in the governing agreement.
A calendar window resets on fixed dates, while a rolling window continuously evaluates the immediately preceding duration. The same incident can therefore affect the two windows differently.
They may. Time-only calculations treat a service as up or down, while request-based SLIs can represent partial impact through failed eligible events. Use the SLI and weighting rules in the service definition.
Possible outcomes include service credits, remediation duties, review, or termination rights, but there is no universal consequence. Check the agreement's measurement, exclusion, claim, and remedy clauses.
