Networking · Reliability
Uptime & SLA Calculator — Allowed Downtime for Any Number of Nines
Convert an availability target such as 99.9 % (three nines) into the downtime it allows each year, quarter, month, week and day, or turn the downtime you actually had into an uptime percentage. It also works out the availability of services chained together or run as redundant copies, so you can see what an SLA really promises before you sign or offer one.
Also listed inDeveloper / IT
Your inputs
Quick start — try a typical case
Pick what you know. With an SLA target, find the downtime it allows. With an outage you had, find the uptime you achieved. With several services that depend on each other, find their combined availability.
The share of time a service promises to be working, from its SLA (service level agreement — the uptime a provider commits to, often with refunds if missed). 99.9 % is "three nines" and allows about 43 minutes of downtime a month; 99.99 % allows about 4 minutes.
Your result Live
Allowed downtime per year
8 h 45 min 57.6 s
525.96 min = 31,557.6 s a year at 99.9% uptime
- Per quarter
- 2 h 11 min 29.4 s (131.49 min)
- Per month
- 43 min 49.8 s (43.83 min)
- Per week
- 10 min 4.8 s (10.08 min)
- Per day
- 1 min 26.4 s (1.44 min)
- Number of nines
- 3 (three nines)
- Downtime share
- 0.1% of the time
Working
Uptime = 99.9%
1: Step 1 · Share of time allowed down
100% − 99.9% = 0.1%
Expression becomes 0.001
2: Step 2 · Seconds in a year
365.25 days × 86,400 s
Expression becomes 31,557,600 s
3: Step 3 · Downtime per year
0.001 × 31,557,600 s
Expression becomes 31,557.6 s = 8 h 45 min 57.6 s
4: Step 4 · Per month (year ÷ 12)
31,557.6 s ÷ 12
Expression becomes 2,629.8 s = 43 min 49.8 s
5: Step 5 · Per week
0.001 × 604,800 s
Expression becomes 604.8 s = 10 min 4.8 s
6: Step 6 · Per day
0.001 × 86,400 s
Expression becomes 86.4 s = 1 min 26.4 s
At 99.9% the service may be down 43 min 49.8 s a month.
The nines ladder — allowed downtime
6 rowsMost SLAs are measured per calendar month, so a bad month can't be made up by a good one — check the per-month figure.
Read the SLA's definition of downtime: planned maintenance is often excluded, and short blips (under 1 to 5 minutes) may not count.
Missing an SLA usually earns service credits (a partial refund), not compensation for your lost business.
A year is taken as 365.25 days (31,557,600 s); a month is a twelfth of that (30.44 days), a quarter a quarter of it.
How this is calculated
Downtime = (1 − A) × period · A = 1 − downtime ÷ period · Nines = −log₁₀(1 − A) · Series: A = A₁ × A₂ × A₃ · Redundant: A = 1 − (1 − A₁)(1 − A₂)(1 − A₃)
- A
- Availability (uptime) as a fraction: 99.9 % = 0.999
- period
- Length of the window in seconds: year = 365.25 × 86,400 = 31,557,600 s; month = year ÷ 12; week = 604,800 s; day = 86,400 s
- 1 − A
- Unavailability: the share of time the service may be down
- A₁, A₂, A₃
- Availabilities of the individual services being combined
An SLA (service level agreement) promises that a service will be up for a set share of the time, written as a percentage such as 99.9 %. The rest is the downtime you have agreed to live with: 0.1 % of a 365.25-day year is 31,557.6 seconds, or 8 h 45 min 57.6 s. People count the 9s — 99.9 % is "three nines", 99.999 % is "five nines" — and each extra nine cuts the allowed downtime tenfold, from about 43 minutes a month to 4 minutes to 26 seconds. When services depend on each other in a chain their availabilities multiply, so two 99.9 % services give only 99.8 %; when they back each other up, the system fails only if all fail at once, so two 99 % servers together reach 99.99 %.
Assumptions & sources
- A year is 365.25 days (31,557,600 seconds) so leap years average out; you can switch to 365 days under advanced options. A month is a twelfth of a year (about 30.44 days), a quarter a quarter of it.
- Downtime is spread over the whole period; real SLAs are usually judged per calendar month, and a 28-day February allows slightly less time than this average month.
- Combined availability assumes services fail independently of each other, and that redundant copies fail over instantly.
- Uptime is treated as all-or-nothing: slow or partly working service counts as up unless your SLA says otherwise.
- 1Site Reliability Engineering — Chapter 3: Embracing Risk — Google (O'Reilly, 2016). Time-based and request-based availability; why 100 % is the wrong target.
- 2Site Reliability Engineering — Appendix A: Availability Table — Google (O'Reilly, 2016). Allowed unavailability per year, quarter, month, week and day for each availability level.
- 3The Site Reliability Workbook — Implementing SLOs — Google (O'Reilly, 2018)
- 4Amazon Compute Service Level Agreement — Amazon Web Services. An example of a monthly uptime percentage commitment with service credits.
Calculator ID NET-UPTIME-001 · version 1.0.0
Common questions
How much downtime does 99.9 % uptime allow?
About 8 h 46 min a year, 43 min 50 s a month, 10 min 5 s a week or 1 min 26 s a day (using a 365.25-day year). That is "three nines" — a common SLA for cloud services and business websites.
What does "five nines" mean?
99.999 % availability: only about 5 min 16 s of downtime a year, or 26 seconds a month. It is the classic target for telephone networks and critical systems, and is very expensive to reach — every component, plus maintenance, must fit inside that budget.
Why does adding redundancy improve uptime so much?
Because an outage now needs every copy to fail at the same moment. Two independent 99 % servers are each down 1 % of the time, but both are down together only 1 % × 1 % = 0.01 % of the time, so together they reach 99.99 %. Shared risks (same power supply, same bad update) reduce this gain in practice.
Why is my overall uptime lower than each service's SLA?
When a request needs several services in a chain — say a web server, a database and a payment gateway — any one failing breaks it, so their availabilities multiply. Two 99.9 % services in series give 99.8001 %, about double the downtime of either alone.
We were down 2 hours this month. What uptime is that?
2 hours out of an average month of 730.5 hours is 0.274 % downtime, so uptime was about 99.73 %. That misses a 99.9 % SLA (which allows about 44 minutes) but meets 99.5 %.
Should I aim for 100 % uptime?
No. Google's SRE book points out that users can't tell 99.999 % from 100 % because their own phone or Wi-Fi is far less reliable, while each extra nine costs much more. Pick a target that matches what your users actually need.