Topic 13.2
PACELC
In one line
CAP's rarely-quoted second half: even without a partition, you choose between latency and consistency (Else Latency vs Consistency).
Think of it like this
Even when everything is working fine (no partition), you still choose between a fast reply that might be slightly outdated (Latency) or a slower reply that's guaranteed fully up to date (Consistency) — like asking a busy friend 'quick answer or the accurate one?'
Key ideas
- 01
PACELC = If Partition: A vs C; Else: L vs C — the full decision space in one acronym.
- 02
Without a partition, a primary+async-replica setup CHOOSES latency: it serves reads from the replica and accepts staleness.
- 03
A synchronous-replica setup chooses consistency: writes wait for the replica to confirm — higher latency.
- 04
Why interviewers love it: almost nobody quotes it, and it makes your answers sound like a practitioner.
- 05
Apply per path: 'writes → consistent' and 'feed reads → latency' are two different PACELC choices in one system.
- 06
English sentence: 'during a partition I favor availability for reads; without one, I accept bounded staleness for latency where the data tolerates it.'
Explain without notes
Which PACELC choice is 'read replicas with 2s lag' — and where does that choice break?
Practice
Write the PACELC sentence for: payments service, feed service, and a counter service.
Trade-offs
- ↔
Consistency everywhere = latency + reduced availability of fresh data; the whole art is choosing per path.
Run it in production
You've designed it. Now build, operate, and break the same idea hands-on in the DevOps courses:
Completion checklist
I can deliver CAP + PACELC in one paragraph that ends with per-path choices.