Downtime
Downtime is any period during which a service is unavailable or not responding correctly to users.
Downtime is the opposite of uptime: the time a service is unreachable, timing out, or returning errors instead of correct responses. It is what SLAs and SLOs are designed to limit, and what monitoring exists to detect quickly. Downtime is usually counted in minutes per month or hours per year, and it maps directly onto an error budget.
Why it matters: downtime is where reliability turns into money and reputation. Lost transactions, broken integrations, support load, and eroded trust all accrue while a service is down. Because the cost is roughly proportional to duration, shaving minutes off each incident has an outsized effect on the total.
A common misconception is that downtime is only the time an outage lasts. The figure that actually hurts is detection-to-recovery: the window from when the failure began to when service is restored. Slow monitoring adds its detection lag directly to every incident — if your tooling takes five minutes to notice, those minutes are downtime your customers felt before anyone was even paged.
Not all downtime is equal. Planned maintenance announced on a status page is treated very differently from an unplanned outage, and partial degradation (one component failing) is often tracked separately from a full outage. Clear definitions up front keep your downtime accounting honest.
Downtime = Total time × (1 − Uptime %) — or simply the sum of all outage durations.
A 99.9% target over a 30-day month (43,200 minutes) permits 43,200 × 0.001 = 43.2 minutes of downtime.
Frequently asked questions
- How is downtime measured?
- An external monitor records when a service stops responding correctly and when it recovers; the difference is the outage duration. Summing every outage over a period gives total downtime, which you can convert to an uptime percentage.
- Does detection lag count as downtime?
- From the customer's perspective, yes. The clock starts when the service actually fails, not when your tooling notices. Faster checks shrink the detection-lag portion of every incident.
- Is planned maintenance counted as downtime?
- It depends on the agreement. Many SLAs exclude announced maintenance windows, but the time is still unavailability for users, so well-run teams schedule it for low-traffic periods and publish it on a status page.
Start monitoring in under a minute
Free forever, no credit card. Upgrade when you need finer intervals.