Mark Korotin
Distributed Systems

Choosing your trade-offs on purpose

Consistency, latency, and complexity are choices. Pragmatic distributed systems match the problem, not the trend.

Working · 10 min read

CAP theorem does not tell you what to build. It tells you what you cannot have. The useful version of that insight is not "pick two" but "pick the right two for your problem, and be explicit about it."

Most distributed-systems failures I have seen were not caused by the wrong choice. They were caused by no choice — by an architecture that drifted into consistency or availability without anyone deciding which one mattered more.

The pragmatic approach is boring but it works: list your invariants. Identify which ones are correctness requirements and which ones are performance requirements. Then choose an architecture that makes the correctness requirements easy to enforce and the performance requirements achievable.