CAP Theorem

In the CAP theorem (introduced by Eric Brewer), a distributed system can only fully guarantee two out of three properties at the same time:

C — Consistency
A — Availability
P — Partition Tolerance

Let's review more in depth.

Overview

Before going through each letter in CAP Theorem, we need to first understand what the "Netword Partition" means. Then we can understand better.

Network Partition

A partition happens when nodes in a distributed system cannot communicate with each other (e.g., network failure, region outage, etc.). Example:

Node A  ❌  Node B

They are alive, but the network between them is broken.

During a Partition

Now suppose a client sends a write request to Node A. But Node A cannot confirm with Node B because the network is broken. The system now has two options:

Option 1: Stay Available (AP behavior)

Node A accepts the write.
Node B does not know about it.
Now the system has inconsistent data.

Option 2: Stay Consistent (CP behavior)

Node A refuses the write (or blocks it).
It waits until communication with Node B is restored.
Data remains consistent. But the system becomes unavailable for that request.

CP System

What happens during a partition? If nodes can’t talk to each other:

The system refuses writes (or even reads).
It waits until a majority (quorum) can agree.
Some clients may get errors or timeouts.

Think of:

Banking systems
Payment processing
Inventory control

Incorrect data is worse than temporary downtime.

AP System

What happens during a partition? If nodes can't talk to each other:

The system continues to accept reads and writes.
Each node operates independently.
Data may become inconsistent across nodes.

Think of:

Social media feeds
Content delivery networks
Recommendation systems

Temporary inconsistency is acceptable when availability is more important.

AC System

AC systems, Availability + Consistency, are theoretically mentioned, but in practice they cannot exist in a distributed system that can experience network partitions. Let's see why.

CAP says: In the presence of a network partition (P), a distributed system must choose either Consistency (C) or Availability (A).

C — all nodes see the same data at the same time
A — every request eventually gets a response
P — system continues to operate even if the network is partitioned

So P is unavoidable in real distributed systems because network failures can always happen.