Skip to content
uencoder

CAP Theorem

The CAP theorem states that under network partition a distributed system can satisfy only one of consistency (C) or availability (A); all three cannot hold at once.

Purpose

To reveal the fundamental trade-off of distributed systems and guide data store selection.

Mechanism

Consistency (all nodes see the same view), availability (every request gets a response), partition tolerance (operation continues despite partitions). Since partitions are unavoidable, the practical choice is between CP (strong consistency) and AP (high availability). CP: ZooKeeper, Spanner; AP: Cassandra, DynamoDB.

Use cases

  • Distributed storage selection
  • Microservice data architecture
  • Multi-region deployment strategy
  • Consistency vs availability trade-offs

Examples

  • Input 
    During a network partition
    Output 
    Choose C (reject inconsistency) or A (allow it)

    CP vs AP

Common errors

Believing all three of CAP can be satisfied simultaneously
Partitions are unavoidable; choose between CP and AP

Related terms

ACID
ACID is the four properties of relational transactions: atomicity (all or nothing), consistency (constraints preserved), isolation (concurrency without interference), and durability (committed data is never lost).
GLOSSARY
BASE
BASE is the weak-consistency NoSQL model: basically available, soft state, eventually consistent — trading strong consistency for availability and scalability.
GLOSSARY
NoSQL
NoSQL is the umbrella term for non-relational databases, including document, key-value, column-family, and graph models, designed for scalability and flexible schemas.
GLOSSARY

References