ARTICLE · Systems
Availability
Availability is a promise about a user's ability to complete a particular action, measured over a particular period.
Availability
YouTube stays disconnected until you choose to load this player. Loading connects to Google/YouTube, which receives your IP address and may use cookies or browser storage.
A friend who answers every message except the urgent ones may be technically responsive and practically unreliable. The same problem appears in software: a homepage can load while checkout, login, or saving a document is unavailable.
Before discussing “high availability,” name the operation that matters. For a guild hall, being available might mean that a signed-in member can read their inbox and send a letter. A health endpoint returning 200 is useful evidence about one boundary, but it cannot alone establish that whole journey.
Extra copies need independent ways to fail#
Two application instances help when one process crashes. They help less if both rely on the same failed database, credential, or network path. Redundancy deserves a dependency map, otherwise the diagram can look resilient while preserving one shared point of failure.
Recovery time matters too. A service that usually restarts quickly may still require a person to notice and repair some failures. Name that person or monitoring mechanism before relying on the restart story.
A practical availability discussion ends with a scenario: this dependency stops, this user action fails, this signal detects it, and this procedure restores it. Opening the World Gate gives you a deployed service on which to ask those questions. The percentage becomes useful after you know what experience it is measuring.
See also