reyhan base

CAP theorem

3 min read

Definition: The CAP theorem says that a distributed system cannot guarantee Consistency, Availability, and Partition Tolerance all at the same time when a network partition occurs. (AWS Documentation)
In practice, because network failures can happen, you usually have to choose whether to favor Consistency (C) or Availability (A) during a partition. (AWS Documentation)

First, understand the 3 letters

  • C — Consistency: Every read gets the latest correct value (or an error).

  • A — Availability: Every request gets a response, even if some nodes can’t communicate.

  • P — Partition Tolerance: The system continues operating even when nodes cannot communicate with each other because of a network failure. (AWS Documentation)

Simple example

Imagine you have two database servers:

        Network
   ┌───────────────┐
   │               │
Server A         Server B
  $100             $100

Now the network connection between them breaks:

Server A    ❌    Server B
  $100             $100

A user changes the balance on Server A to $50.

Now the system has a problem:

Option 1 — Consistency

Server B cannot confirm the new value, so the system refuses/blocks the operation.

User → Server A → "Sorry, cannot complete"

You preserve C, but sacrifice A during the partition.

Option 2 — Availability

Server A accepts the change and responds immediately:

User → Server A → "Success"

But Server B might still think the balance is $100.

You preserve A, but sacrifice C during the partition.

That’s the core idea of CAP. (AWS Documentation)

3 practical examples

1. Banking system — favor Consistency

If your bank’s servers cannot communicate, you’d rather temporarily reject a transaction than allow two servers to show different account balances.

Network failure
      ↓
Reject transaction
      ↓
Keep balance correct

2. Social media — favor Availability

If Instagram-like servers can’t communicate, you may still want users to post and view content, even if some information temporarily isn’t synchronized.

Network failure
      ↓
Keep serving users
      ↓
Synchronize later

3. Shopping cart — favor Availability

If you add an item to a shopping cart while one server is temporarily disconnected, the system may accept the change and synchronize it later rather than completely preventing you from using the shop.

CP vs AP

This is where you’ll often hear:

CP — Consistency + Partition Tolerance

Network partition
       ↓
Can't guarantee consistency
       ↓
Reject/delay some requests

AP — Availability + Partition Tolerance

Network partition
       ↓
Keep accepting requests
       ↓
Data may temporarily differ

You will sometimes hear people say “pick any two of C, A, and P.” That’s a useful beginner mnemonic, but technically the important trade-off happens when a partition occurs: you generally choose between consistency and availability. (AWS Documentation)

⚠️ One important clarification

CAP’s Consistency is not the same thing as the C in ACID. In the formal CAP definition, consistency refers specifically to a strong guarantee called linearizability. (Martin Kleppmann)

Easy way to remember:

C: “Give me the correct/latest answer.”
A: “Give me an answer.”
P: “Keep working even when servers can’t talk to each other.”

Back-End My-Journey-In-Codeless

Graph View