CAP Theorem Deep-Dive and BASE vs ACID

Q: CAP Theorem Deep-Dive and BASE vs ACID

**CAP** states that during a network partition, a distributed system can guarantee at most two of Consistency, Availability, and Partition tolerance — and since partitions are unavoidable, the real choice is **CP or AP during partition**. **ACID** (Atomicity, Consistency, Isolation, Durability) gives transactional guarantees ideal for money movement; **BASE** (Basically Available, Soft state, Eventually consistent) trades those guarantees for scale and availability. Modern systems mix both — ACID for orders/payments, BASE for counters/caches. MongoDB 4.0+ adds multi-document transactions, DynamoDB supports `TransactWriteItems` up to 100 items, and across services the **saga pattern** with outbox + compensating actions replaces distributed two-phase commit.

Intermediate·2 min read·

AmazonMongodbStripeUberNetflix

Fast recall

Quick revision

Core answer

CAP states that during a network partition, a distributed system can guarantee at most two of Consistency, Availability, and Partition tolerance — and since partitions are unavoidable, the real choice is CP or AP during partition. ACID (Atomicity, Consistency, Isolation, Durability) gives transactional guarantees ideal for money movement; BASE (Basically Available, Soft state, Eventually consistent) trades those guarantees for scale and availability. Modern systems mix both — ACID for orders/payments, BASE for counters/caches. MongoDB 4.0+ adds multi-document transactions, DynamoDB supports TransactWriteItems up to 100 items, and across services the saga pattern with outbox + compensating actions replaces distributed two-phase commit.

Interview-ready guide

Interview answer

Define each precisely

CAP: during a partition, pick C or A; P is mandatory
ACID: single-node transaction contract — atomicity, consistency, isolation, durability
BASE: distributed philosophy — basically available, soft state, eventually consistent

Clarify the C confusion

CAP's C is linearizability across replicas
ACID's C is constraint preservation within a transaction
Different concepts, same letter

Give concrete stores

ACID strong: Postgres, MySQL, Spanner, MongoDB with majority concerns
BASE default: Cassandra, DynamoDB, Riak
Tunable: most modern NoSQL

Mention sagas

Distributed transactions across services use sagas, not 2PC
Compensating actions + outbox pattern + idempotence
Accepts eventual consistency for operational simplicity

03 · Deep dive