Captheorema

CAPtheorema, commonly known as the CAP theorem or Brewer's theorem, is a principle in distributed computing that describes trade-offs among three properties: Consistency, Availability, and Partition tolerance. The theorem asserts that in a distributed data store that tolerates network partitions, it is impossible to guarantee all three properties simultaneously. Therefore, a system must prioritize two of the three, accepting the compromise on the third.

Definitions:

- Consistency means that all reads reflect the most recent write or an error; every operation appears to happen in a single, coherent system order.

- Availability means that every request receives a (non-error) response, without guaranteed freshness of data.

- Partition tolerance means the system continues to operate despite message loss or delays between nodes, effectively managing network partitions.

Historical context:

The CAP theorem was proposed by Eric Brewer in 2000 and later formalized and proven by Seth Gilbert and Nancy Lynch in 2002. It is a theoretical framework for understanding the limitations of distributed systems and is frequently cited in discussions of database design and distributed architectures. It is often misunderstood as prescribing fixed classifications, but it is more accurately described as a trade-off framework applicable in the presence of partitions.

Practical implications:

In practice, systems are described as CP (consistency and partition tolerance), prioritizing data correctness over availability during partitions; or AP (availability and partition tolerance), prioritizing responsiveness with eventual consistency; since partition tolerance is essential, CA (consistency and availability without partition tolerance) is not achievable in a distributed system that can experience partitions. Modern designs commonly use quorum protocols, Paxos or Raft for CP goals, or Dynamo-style approaches for AP goals, often with tunable consistency.