Case Study: Monolith-to-Microservices Migration

Most companies that run microservices did not start there. They started with a single application that worked well for years, grew a large team and codebase around it, and only then began to feel it slowing them down. The interesting question is almost never "monolith or microservices" as a clean greenfield choice. It is "we have a monolith that is starting to hurt, and a business that cannot stop while we fix it, so how do we get from here to there without breaking anything?" That is a multi-year engineering and organizational problem.

To make the journey concrete, this chapter follows a composite migration: a Rails monolith serving a few million users for an online marketplace, the kind of system Etsy and Shopify ran for years before deciding how far to decompose. The numbers and timeline are illustrative rather than from one company, because the shape of these migrations is remarkably consistent.

What matters is the sequence of phases, the decisions at each one, and the realistic horizon. A real migration of a mature system takes three to five years, and understanding why it takes that long is most of what separates a credible plan from a naive one.