Skip to content

What is downtime?

Downtime is any period when a system fails to do the job it exists for: the website that won't load, the server that doesn't answer, the API that returns errors instead of data. Uptime and downtime are the same measurement read from opposite ends — a month at 99.9% uptime is a month with about 43 minutes of downtime.

What counts as down

From the outside — which is where your clients sit — a system is down when a request to it fails. That covers three shapes of failure:

  • No answer. The host is unreachable: crashed, powered off, or cut off by the network.
  • A wrong answer. The server responds, but with an error — a 5xx status from a web application, a refused connection on a service port.
  • A very late answer. The system technically responds, but slower than anyone will wait. Monitoring draws a hard line here: UpAll's checks time out at a fixed 10 seconds, and past that the check fails.

A response that's slow but arrives within the limit is a different state — not down but degraded; see what is latency.

Monitoring turns a fuzzy word into a precise span: downtime starts at the first failed check of a streak and ends at the next successful one. UpAll opens an incident after a configurable number of consecutive failures and backdates its start to that first failure, so the recorded downtime matches what actually happened — see how incidents open and close.

Common causes

  • Hardware. Disks die, power supplies fail, machines overheat. Rarer in the cloud era, still routine in on-premise server rooms.
  • Deployments and configuration changes. The largest self-inflicted category: a bad release, a migration that locks a table, a firewall rule that cut more than intended. Outages cluster suspiciously close to moments when someone changed something.
  • DNS. An expired domain or a wrong record change makes a perfectly healthy system unreachable.
  • Certificates. An expired TLS certificate makes browsers refuse the connection — indistinguishable from an outage to the user. See certificate expiry risks.
  • Upstream providers. The hosting company, the client's ISP, a cloud region, a third-party API. Not your fault, still your problem — and your monitoring sees it first.

Planned vs unplanned

Planned downtime is scheduled and announced: maintenance, upgrades, migrations. Unplanned downtime is everything that surprises you. Contracts usually treat them differently — an agreed maintenance window typically doesn't count against an SLA target.

UpAll has no separate scheduled-maintenance feature. The practical approach during planned work is to pause the monitor — checks stop, so no incident opens and no alert emails go out — then re-enable it when you finish. See pausing monitors.

What downtime costs

For a small or mid-size business the bill has three lines:

  • Direct revenue. An online store averaging 40 orders an hour at 150 SAR each loses about 6,000 SAR for every down hour — before counting the customers who tried once and never came back.
  • Idle staff. When the ERP or the point-of-sale system is down, salaried employees stand around waiting for it.
  • Trust. The slowest cost to repair. Clients rarely remember the eleven flawless months; they remember the morning nothing worked.

For an MSP the costs compound: every minute of a client's downtime is your phone ringing and your credibility on the line. That's exactly why downtime is worth measuring precisely — minutes are what it costs, and minutes are what monitoring counts. For how those minutes map to availability targets, see the nines of availability.