Why 'Choose Two' is Oversimplified
You'll learn: Why the popular "pick two out of three" interpretation of CAP Theorem misses how distributed systems actually behave.
The Myth of Binary Choices
You've probably heard CAP summarized as "pick two: Consistency, Availability, or Partition tolerance." This makes it sound like a permanent, all-or-nothing decision—like choosing toppings on a pizza. But that's not how real systems work.
The truth? Partitions are rare events, not constant states. Your system doesn't operate in partition mode 24/7. Most of the time, the network is healthy, and you can have both consistency and availability simultaneously. The CAP tradeoff only matters during the partition itself.
Normal Operation vs Partition Scenarios
Think of it like a fire escape plan: you don't choose between "use stairs" or "use elevator" for your entire building's lifetime. During normal operation, you use elevators freely. Only during a fire (partition) do you make the tradeoff and take the stairs.
Similarly:
- Normal operation: Your distributed database can be both consistent and available
- During partition: Now you must choose—wait for consistency (CP) or serve potentially stale data (AP)
The "choose two" framing makes it seem like you're building a CP system or an AP system forever. In reality, you're choosing how to behave when partitions occur, not how to behave always.
Beyond Simple Labels
Modern systems often use hybrid strategies: strong consistency for critical data (user balances), eventual consistency for less critical data (profile views). You're not locked into one camp.
Key Takeaway: CAP isn't about permanently sacrificing one property—it's about deciding your behavior during rare partition events while enjoying all three properties during normal operation.