Definition

The CAP theorem (Brewer’s theorem) states that in the presence of a network partition, a distributed data store can only guarantee two of the following three: Consistency, Availability, and Partition Tolerance.

Consistency (all nodes see the same data at the same time).

Availability (every request receives a response, either success or failure).

Partition Tolerance (system continues to operate despite network failures between nodes).

Core concept

Under a network partition (P), you must choose between consistency and availability. For example, if a cluster splits, you can either allow writes on both sides (remain available but risk inconsistency) or refuse requests on some nodes (remain consistent but sacrifice availability).

Use cases

This theorem guides database design choices. Traditional RDBMS (like single-node DBs) prioritize C/A (no partitions). Distributed NoSQL often choose P/A (partition-tolerant and available, at cost of consistency during partitions), or C/P (sacrificing availability under partitions).

Trade-offs

CAP is about worst-case (partitions). Real systems try to maximize all three normally but must degrade gracefully. The theorem implies you cannot have a perfectly consistent and available system under partitions.

Example

During a network split, a CP system like MongoDB (with primary/replica model and majority writes) might become unavailable on one side to avoid inconsistency; an AP system like Dynamo might allow writes on both sides and reconcile later.

References

“CAP theorem states any distributed data store can provide at most two of: consistency, availability, partition tolerance”[[14]](https://en.wikipedia.org/wiki/CAP_theorem "CAP theorem - Wikipedia").