CAP Theorem — Part 3: CAP in Practice — Real Systems & Trade-offs

What is this?

Parts 1 and 2 gave you the theory. This is where the rubber meets the road: what do the databases on your résumé actually do the moment a network partition hits? Because here's the uncomfortable truth — every system in this chapter made the dinner-planner's choice from Part 1, baked it into its architecture, and lives with the consequences. Spanner refuses to leave the house. Cassandra picks a restaurant and goes. Kafka lets you decide with a config flag.

Knowing which choice each system made — and crucially, how it behaves while partitioned — is the difference between "I've heard of Cassandra" and "I know Cassandra is AP, so I'd never put a bank ledger on it." That distinction is what interviewers are probing for, and it's what keeps you from a 3am page when a single availability zone blips.

The one-line idea: every real distributed database has already chosen CP or AP and wired it into its bones — this chapter maps the major systems onto that choice so you can pick the right tool and predict its behavior during the failure that will eventually come.