"We guarantee 99.9% uptime" sounds almost perfect, until you see what it allows. The percentage hides how much time is actually lost, and the difference between each extra "nine" is large.
The downtime each percentage allows
The table shows the maximum downtime, if you are down for exactly that share of the time.
| Uptime | Per day | Per week | Per month (30 days) | Per year |
|---|---|---|---|---|
| 99% | 14.4 minutes | 1.68 hours | 7.2 hours | 87.6 hours (3.65 days) |
| 99.5% | 7.2 minutes | 50.4 minutes | 3.6 hours | 43.8 hours |
| 99.9% | 1.44 minutes | 10.1 minutes | 43.2 minutes | 8.76 hours |
| 99.95% | 43 seconds | 5 minutes | 21.6 minutes | 4.38 hours |
| 99.99% | 8.6 seconds | 1 minute | 4.3 minutes | 52.6 minutes |
| 99.999% | 0.9 seconds | 6 seconds | 26 seconds | 5.3 minutes |
So 99% sounds high but allows more than three and a half days of downtime a year, while 99.9% allows under nine hours. Each extra nine divides the allowed downtime by ten, and typically multiplies the cost and effort of delivering it.
How is uptime measured?
Two sites can both claim 99.9% and mean very different things. The questions to ask:
- What counts as down? Only a completely unreachable server, or also error pages and extreme slowness?
- From where is it measured? A check from one location can miss a problem that affects only some visitors.
- How often is it checked? If you check every five minutes, an outage of three minutes may never be seen.
- Is planned maintenance excluded? Many service guarantees do not count scheduled maintenance.
- Over what period? A yearly figure can hide a terrible month.
Choosing a realistic target
- A personal site or a small blog: 99% to 99.5% is perfectly reasonable, and a few hours of downtime a month is rarely a disaster.
- A small business site or a shop: aim for 99.9%. At that level, any unplanned outage is something to investigate.
- A business where minutes matter, such as payments or a critical application: 99.95% or 99.99% takes redundancy and operational discipline, and the monitoring should be as strict as the target.
Pick the figure from the cost of downtime, not from what sounds impressive. If an hour offline costs you little, spending heavily to avoid it makes no sense. If it costs you thousands, a cheap hosting plan with no monitoring is the wrong choice.
What you can control
- Monitor from outside, every minute, so short outages are counted and you hear about them immediately. See what to monitor on your website.
- Prevent the avoidable outages. A large share of downtime comes from expired certificates and domains, which are completely predictable. See SSL certificate expired: how to fix it and how to check when a domain expires.
- Plan maintenance for quiet hours, and mark it as planned so it does not count against you or flood your alerts.
- Measure honestly. Track the percentage over time, per month, and look at the worst month, not only the average.
ServerMonitor shows 24-hour, 7-day and 30-day uptime for every site, and keeps the history for up to a year on paid plans, so you can see your real number instead of guessing. Planned downtime can be left out of the figure when you use maintenance windows.