CAP Theorem — Interview Cheatsheet#

Step 1: What consistency does this data need?
  Financial, locks, leader election → linearizability → CP
  Social, analytics, preferences    → eventual/causal → AP

Step 2: What happens if this system is unavailable?
  Users lose money, critical infra breaks → tolerate unavailability (CP)
  Users see stale feed, minor annoyance   → tolerate staleness (AP)

Step 3: Which is worse — wrong data or no response?
  Wrong data worse  → CP
  No response worse → AP

Step 4: State the choice and justify it
  "This system needs CP because stale [X] would cause [Y]"
  "This system needs AP because unavailability would cause [Y]"

Step 1: Financial data — needs linearizability
        Stale balance → charge against non-existent funds → financial loss

Step 2: If unavailable → user sees "try again" → calls support → recoverable

Step 3: Wrong data (double charge, missing debit) is catastrophic
        No response is recoverable

Step 4: "This needs CP. A failed transaction is recoverable.
         A wrong transaction destroys trust and violates regulations."

DB choice: PostgreSQL with SERIALIZABLE isolation
           Or Google Spanner (globally consistent)

Step 1: Message delivery — needs availability, not linearizability
        Slight reordering or eventual delivery is acceptable

Step 2: If unavailable → 2 billion users can't send messages → unacceptable

Step 3: Slightly delayed message delivery is annoying
        2 billion users unable to communicate is catastrophic

Step 4: "This needs AP. Availability is non-negotiable at this scale.
         We accept eventual consistency — messages deliver, maybe slightly out of order."

DB choice: Cassandra (AP, eventually consistent, massive write throughput)

Step 1: Inventory (room availability) needs strong consistency
        Stale room count → two users book same room → double booking

Step 2: If unavailable → user retries → minor inconvenience

Step 3: Double booking = hotel ops problem, customer service nightmare
        Unavailability = user tries again in 10 seconds

Step 4: "This needs CP. Double booking is a serious business problem.
         Temporary unavailability is acceptable."

DB choice: PostgreSQL with SELECT FOR UPDATE or SERIALIZABLE

Wrong answer: "We'll sacrifice partition tolerance"
Right answer: "CA doesn't exist in distributed systems.
               Partitions will happen — the real choice is CP or AP."

Wrong answer: "C means the data is consistent"
Right answer: "C in CAP specifically means linearizability —
               every read sees the latest write globally.
               AP systems still have eventual consistency — they just
               can't guarantee linearizability during a partition."

Wrong answer: "CAP governs the whole system"
Right answer: "CAP only describes behavior DURING a network partition.
               During normal operation, you can have both C and A.
               The choice matters for how you behave in failure cases."

Cassandra with ONE   → AP (available, potentially stale)
Cassandra with ALL   → CP (consistent, may be unavailable)
DynamoDB eventually consistent reads → AP
DynamoDB strongly consistent reads   → CP