Hosting Uptime: Read the Percentage and Test the User Journey
A server can be running while your website’s checkout fails. Before comparing uptime percentages, ask what was measured, where the check ran and what counted as success.
There is no useful percentage without a measurement window and a definition. A provider’s infrastructure status, an external homepage check and a completed purchase measure different things.
Calculate a time-based example carefully
For a defined service and observation period, time-based uptime is available time divided by total observed time, multiplied by 100. Whether maintenance or partial failure is included depends on the chosen measurement rules.
| Illustrative target | Unavailable time in a 30-day window |
|---|---|
| 99% | 7 hours 12 minutes |
| 99.9% | 43 minutes 12 seconds |
| 99.99% | 4 minutes 19.2 seconds |
These figures use a continuous 43,200-minute window and simple arithmetic. They are not measurements of any host, an outage schedule or an interpretation of a particular service agreement.
Distinguish running from working
A machine’s operating-system uptime says how long it has run since a restart. It does not establish that DNS, HTTPS, the database or an external dependency has worked throughout that period.
A homepage returning HTTP 200 can still contain an error message. Set checks for the behaviour visitors need, not just a reachable port. For a store, that may include product search and a safe checkout test using the payment provider’s supported test mode.
Request success can reveal partial failure
For some services, the share of valid requests completed successfully is more informative than a binary up-or-down clock. Google’s SRE discussion of reliability describes this approach.
Define eligible requests and success criteria before collecting results. Also measure response time: a technically successful request that arrives too late may still fail the user’s task.
Read the service promise in its own terms
A target is not proof of past performance. Check the covered product, calculation window, measurement method, exclusions and the supplier’s incident process in the current service documentation.
If a credit process is offered, establish what evidence and reporting steps it requires. Do not treat a possible credit as recovery of lost orders or as a substitute for your own recovery plan.
Monitor from outside the hosting account
Use appropriate external checks from the locations that matter to your audience. Record timestamps, failing steps, response codes and relevant application logs. A local monitoring failure needs investigation before it is labelled a global outage.
Send alerts to someone who can act. Our support evaluation guide helps establish escalation routes, while the security and recovery checklist covers restore evidence.
Choose a reliability target for the actual site
Consider which tasks can fail, how quickly they must recover and what failure costs your operation. A cloud label or higher hosting tier does not automatically supply application failover.
Use the hosting worksheet to record the target, checks and recovery responsibilities. Keep measured observations separate from marketing promises.








