The Hidden Cost of Chatty Microservices

By Marta Ellison · July 20, 2026 · Engineering

Splitting a monolith into services trades local complexity for distributed complexity, and nowhere is that clearer than in fan-out. A request that used to make three function calls now makes three network calls, each with its own timeout, retry policy, and failure mode.

The arithmetic is unforgiving. If each hop has a 99.9% success rate and your request touches ten services, end-to-end success drops below 99%. Users do not experience your architecture diagram - they experience the product of every dependency's reliability.

Before splitting anything, draw the call graph of your hottest request and count the network edges. Every edge you remove is worth more than any retry tuning you will ever do.