SLA availability and downtime calculator
How many hours of downtime is 99.9% a year? Does the downtime you measured meet the SLA? What is the combined availability of the components in your chain? Enter a percentage, a duration or components and see the result at once. The calculation runs in your browser.
Calculated in your browser; nothing is sent.
Enter it as a percentage; a dot or a comma works as the decimal mark (e.g. 99.95).
Presets
Allowed downtime
99.9%
3.0 nines
- Yearly
- 8 h 45 min 36 s
- Monthly
- 43 min 12 s
- Weekly
- 10 min 5 s
- Daily
- 1 min 26 s
The monthly value follows the month definition you selected.
Common values
| Availability | Nines | Yearly | Monthly | Weekly | Daily |
|---|---|---|---|---|---|
| 99% | 2.0 | 3 d 15 h 36 min | 7 h 12 min | 1 h 40 min 48 s | 14 min 24 s |
| 99.5% | 2.3 | 1 d 19 h 48 min | 3 h 36 min | 50 min 24 s | 7 min 12 s |
| 99.9% | 3.0 | 8 h 45 min 36 s | 43 min 12 s | 10 min 5 s | 1 min 26 s |
| 99.95% | 3.3 | 4 h 22 min 48 s | 21 min 36 s | 5 min 2 s | 43.2 s |
| 99.99% | 4.0 | 52 min 34 s | 4 min 19 s | 1 min | 8.64 s |
| 100% | 5.0 | 5 min 15 s | 25.92 s | 6.05 s | 0.86 s |
Formulas and assumptions
- Allowed downtime = (1 − A) × window. Year = 365 days (8760 h); month = 30 days (720 h) or year ÷ 12 (730 h); week = 168 h; day = 24 h.
- Achieved A = (window − maintenance − downtime) ÷ (window − maintenance). Nines = −log10(1 − A).
- Series: A = A₁ × A₂ × … ; parallel: A = 1 − (1 − A₁)(1 − A₂)… ; k identical copies: 1 − (1 − A₁)^k.
- Component failures are assumed independent.
Window: availability is always defined over a measurement window (calendar month, rolling 30 days, a year). Your contract should state the window, what counts as downtime (partial outage, measurement interval) and whether planned maintenance is excluded; this tool does not interpret these for you.
Results are a preliminary calculation, not a contract interpretation or legal advice. The measurement window, downtime definition, maintenance exclusions and service credit terms in your SLA apply. Real availability depends on your measurement method (for example the check interval and partial outages).
Shall we review your service level targets, your redundancy design and your monitoring together?
Request a meeting01
How to use
A
For allowed downtime, enter an availability percentage or pick a preset; see the yearly, monthly, weekly and daily time.
B
For achieved availability, enter the window, the total downtime you measured (and any maintenance outside the SLA); optionally add your target to see the downtime budget.
C
For a chain, write each component as name;percent, add a copy count for redundant ones and choose series or parallel; see the combined value and the weakest link.
02
Availability percentage and downtime
Availability is the share of a window during which the service is working. Allowed downtime is the remaining share of the window: for 99.9% about 8.76 hours a year (8 h 45 min) and about 43 minutes a month; for 99.99% about 52.6 minutes a year; for 99.999% about 5 minutes 15 seconds a year. Each extra "nine" cuts the allowed downtime to a tenth.
The month definition affects the result: a 30-day month is 720 hours, year ÷ 12 is 730 hours. The difference is small, but you need to know which one your contract uses. The tool's table and results let you choose either.
03
Series and parallel components
If a service depends on many components and all of them must work (power, network, server, storage, application), the availabilities multiply: four components at 99.9% give a combined value of about 99.6%. The chain is always lower than its weakest link, so for a 99.99% target every component must sit far above the target or be built redundantly.
With redundant components the service fails only if all fail at once: two independent 99.9% copies in parallel give 99.9999%. That is an upper bound, though: if the copies depend on the same power feed, the same software or the same maintenance window, a common-cause failure reduces the gain. The tool assumes independence; the real value may be lower than calculated.
04
Measurement window and maintenance exclusion
The same outage gives a different percentage depending on the window: a one-hour outage exceeds the 99.99% budget over a year (about 52.6 minutes), while over a month it corresponds to about 99.86%. Without knowing whether your SLA uses a calendar month, a rolling 30 days or a year, results cannot be compared.
Many SLAs deduct planned maintenance. The tool does this by shortening the effective window: availability is calculated over (window − maintenance). The contract should state what counts as planned maintenance (notice period, maximum duration) and how partial outages are counted. This tool is not a contract interpretation; follow the definition in your SLA.
FAQ
- Are the values I enter sent anywhere?
- No. All calculation runs in code in your browser; what you type is not sent to any server and is not stored.
- How much downtime a year is 99.9%?
- 8.76 hours a year, or about 8 hours 45 minutes 36 seconds (based on 365 days). That is about 43 minutes 12 seconds a month (30-day month).
- Should I use 30 days or 365/12 days for a month?
- Use your contract's definition. In SLAs based on the calendar month, the actual number of days applies (28–31). The tool's two options show the two common approaches so you can see the difference.
- Is the maintenance window part of the SLA?
- It depends on the contract. Many SLAs exclude announced planned maintenance. If it is excluded, enter the maintenance hours in 'Achieved availability' mode and the window shrinks accordingly. If not, leave 0.
- Why do two redundant servers not give exactly twice the reliability?
- Redundancy cuts unavailability exponentially (u²), but only if failures are independent. A shared power feed, network, software bug or a wrong configuration can take both down; failover time also counts as downtime.
Back your SLA target with architecture
Let us design redundancy, monitoring and recovery for the availability you are aiming at. In a free discovery call we discuss your current architecture and service level targets.