Explain the CAP theorem and what it means when choosing a distributed datastore.

Question

Accepted Answer

CAP is about 3 properties of a distributed datastore: consistency, every read sees the most recent write; availability, every request gets a non-error response; and partition tolerance, the system keeps functioning when the network between nodes splits. The popular phrasing is pick 2, but that is misleading. In any real distributed system partitions are a fact of life, you will get dropped packets and split nodes, so partition tolerance is not something you can give up. The real choice happens during a partition, and it is consistency versus availability: you cannot have both while the network is split. A CP system refuses or blocks requests it cannot serve consistently, it would rather return errors for 30 seconds while a leader election completes than hand back stale or conflicting data. An AP system keeps responding, accepting that some nodes serve data that is a few seconds behind, and reconciles the divergence afterwards, eventual consistency. So when I pick a datastore I map it to what the data actually needs: for account balances, inventory counts, anything where wrong data is worse than a brief outage, I lean CP; for a social feed, product view counts, or a shopping cart, where staying available matters more and eventual consistency is fine, I lean AP. I would also mention PACELC as the more complete model, because even with no partition there is still a latency versus consistency trade-off, and that one you pay on every single request. The point is to reason about the specific data, not recite the triangle.

Answer

Why is "pick two" a misleading way to state CAP?

Partition tolerance is not optional in a distributed system, partitions will happen, so you do not really get to drop P. The genuine choice is C vs A during a partition; the rest of the time you have both.

Answer

Give an example of when you would pick CP over AP.

Data where serving stale/conflicting values is worse than being briefly unavailable, account balances, inventory, bookings. You would rather reject a request during a partition than double-sell or show a wrong balance.

Explain the CAP theorem and what it means when choosing a distributed datastore.

Explain the CAP theorem and what it means when choosing a distributed datastore.

The follow-ups they ask next

What the interviewer is listening for

What sinks the answer

If you genuinely do not know

Keep going with databases

Knowing the answer is not the same as recalling it under pressure