Pick two: consistency, availability, partition tolerance
Day 124 of 149
👉 Full deep-dive with code examples
The Three Wishes Analogy
A genie says: "Pick any TWO wishes, but not all three"
- Wish 1: Fastest car
- Wish 2: Cheapest car
- Wish 3: Most reliable car
You can have fast + cheap (less reliable), or cheap + reliable (not fast).
CAP is like this—but the real trade-off shows up when the network is split (a partition).
The Three Properties
C - Consistency:
- Reads behave as if there were a single up-to-date copy (strong consistency)
- In practice, “consistency” has levels; CAP discussions usually mean a very strong form
A - Availability:
- Every request gets a response from a non-failing node
- That response might be an error (the key point is: the system doesn’t stop responding)
P - Partition Tolerance:
- The system keeps operating even if messages between servers are dropped/delayed (a network partition)
Why You Can't Have All Three
Imagine two servers that should stay in sync:
Server A ←─ network break ─→ Server B
Network breaks. A user writes to Server A.
Option 1: Stay Consistent
- Don't let Server B respond until synced
- But Server B becomes unavailable!
Option 2: Stay Available
- Both servers respond with what they have
- But data is now inconsistent!
During a partition, you must choose what to prioritize.
Real-World Choices
CP (Consistency + Partition Tolerance):
- Bank transactions
- Might be briefly unavailable
AP (Availability + Partition Tolerance):
- Social media feeds
- Might show slightly stale data
CA (rarely used):
- Only works when you assume no partitions (for example: a single-node system or a very reliable network)
In One Sentence
CAP Theorem says that when a network partition happens, a distributed system has to trade off between consistency and availability.
🔗 Enjoying these? Follow for daily ELI5 explanations!
Making complex tech concepts simple, one day at a time.
Top comments (0)