What Does CAP Mean?
CAP describes a limit in distributed systems: during a network partition, a system cannot guarantee both linearizable consistency and availability for every request.
โข
Consistency: operations behave as if there were one up-to-date copy, respecting the order of completed operations.
โข
Availability: every request to a non-failing node eventually receives a result, rather than being left waiting indefinitely.
โข
Partition: nodes cannot reliably exchange messages with one another.
Example: Two Copies of a Balance
Suppose two nodes lose contact and one receives a balance update. The other cannot know whether its copy is current. Refusing or delaying some operations can protect consistency. Continuing to accept operations on both sides can preserve availability, but the system needs rules for stale reads or conflicting updates.
Common Misunderstandings
โPick any twoโ hides the important condition: the conflict arises when a partition occurs. Partition tolerance is not simply an optional feature that makes unreliable networks disappear.
CAP consistency is not the same as the C in ACID, which concerns application invariants. SQL and NoSQL are also not fixed CAP categories. Evaluate a productโs deployment, operation, and consistency settings rather than assigning guarantees from its label alone.
