Art of Codeartofcode
Lesson 1 of 17

16. Consistency and CAP

1
The picture to keep in your head
2
16.1 What does "consistent" actually mean?
3
16.2 What is strong consistency?
4
16.3 What is eventual consistency?
5
16.4 What are read-your-writes and monotonic reads?
6
16.5 What is a network partition?
7
16.6 What does the CAP theorem actually say?
8
16.7 What do CP and AP systems look like in practice?
9
16.8 What is PACELC?
10
16.9 How do quorums give tunable consistency?
11
16.10 How do you choose consistency for each feature?
12
Memory card and where the analogy breaks
13
Practice: five levels
14
Interview questions
15
Solutions
16
Sources
17
Code lab
16. Consistency and CAP

The picture to keep in your head

3 min read+10 XP

Every question below is answered twice: first at the bank counters, so you can picture it, then in the system, so you know exactly how it works. Each answer ends with a memory code to lock it in.

The picture to keep in your head

A bank has two counters in different towns, both serving the same account. Normally the clerks phone each other before every withdrawal: "Asha wants ₹400; is her balance still ₹500?" Then one day the phone line is cut. Asha walks up to counter 2 and asks for ₹400. The clerk has a choice. Refuse ("come back when the line is fixed"), which is safe but leaves Asha without cash. Or pay out from the last balance they knew, which keeps Asha happy but risks paying money that was already withdrawn at counter 1.

And even on a normal day, phoning first makes every withdrawal slower.

That's the whole of this chapter: consistency (do all copies agree?), what happens when copies can't talk (CAP), and what consistency costs on a normal day (PACELC).

Our running example: the bookstore's gift-card wallet, now replicated in Mumbai and Frankfurt (Chapter 14). A customer can spend their gift-card balance from either region.

At the bankIn the system
Two counters, one accountTwo replicas of the same data
Phoning before every withdrawalCoordinating replicas before answering
The line being cutA network partition
"Come back later"Choosing consistency: return an error rather than a wrong answer
Paying from the last known balanceChoosing availability: answer, maybe stale
Every withdrawal taking longer because of the callThe latency cost of consistency (PACELC)

two bank counters mapped to replicas, partitions, CP, AP and latency.

Figure: when the line is cut, each counter must choose: refuse the customer, or risk a wrong answer.