Uptime monitoring
Uptime monitoring is the automated checking of whether a website, application or server is available and responding as expected. It helps detect outages, trigger alerts and track availability over time.
How uptime monitoring works
A monitor sends scheduled requests to your service. It checks the response against rules such as an expected status code, response content and a timeout. Failed checks trigger alerts according to your configuration.
Run checks outside your own infrastructure so an outage cannot silence the monitor. Test what customers need, not just whether the homepage loads: a successful HTTP response does not prove that login or checkout works.
Check frequency matters. A short outage can start and end between checks, so periodic monitoring estimates availability rather than recording every second of downtime.
What 99.9% uptime allows over 30 days
A 30-day month contains 30 × 24 × 60 = 43,200 minutes. With time-based measurement and no excluded maintenance:
| Availability target | Total downtime budget |
|---|---|
| 99.9% | 43.2 minutes |
| 99.99% | 4.32 minutes |
The calculation for 99.9% is:
43,200 × (1 − 0.999) = 43.2 minutes
That budget covers all outages combined, not each incident. Actual time-based uptime is calculated as:
(Total time − downtime) ÷ total time × 100.
A dashboard percentage still depends on what was checked. Monitoring one endpoint does not establish availability for every customer or feature.
Questions people ask
Is 99.9% uptime good?
It depends on the service. About 43 minutes of downtime in a 30-day month might be acceptable for an informational website but costly for software customers need to run their business. For a startup, choose a target based on customer commitments and outage costs, not a status-page badge.
Which website uptime monitoring tool is the best?
There is no universal winner. Choose one that checks your critical endpoints from outside your infrastructure, validates responses beyond status codes and alerts someone who can act. Compare check frequency, monitoring locations and alert delivery before price. A dashboard nobody watches is not an incident response plan.
