Distributed System Trade-offs (CAP & Beyond)

Back to Distributed Systems
Arjit Sharma
01 Min

CAP Theorem

image.png

It is impossible for a distributed data store to guarantee all three of the following properties -

  • Consisteny (C) - Every read gets latest write.
  • Availability (A) - Every request gets a response.
  • Partial Tolerance (P) - System continues working despite network failures.

Or more correctly -

“ A distributed system can be either consistent or available in the presence of a network partition. “

We prioritize Consistency in Banking transactions, Stock trading systems, while for Social media feeds and E-commerce carts, Availability is prioritized.

Note - ACID focusses on ensuring integrity and reliability of individual transaction in system while CAP deals with broader trade-offs across distributed system

PACELC theorem

The PACELC theorem extends CAP by addressing what happens when there is no partition:

  • If a partition occurs (P), the system must choose between Availability (A) and Consistency (C).
  • Else (E), when there is no partition, the system must choose between Latency (L) and Consistency (C).

Example - Amazon DynamoDB is PA/EL (chooses availability during partitions but optimizes for low latency when there are no partitions. Google Spanner is PC/EC (chooses Consistency in all cases)

Next →