Cloud-Native Migration

The failure mode of a migration is not technical. It is that the business stops shipping for eighteen months, loses patience, and the half-finished migration becomes permanent — now you maintain two architectures instead of one.

So the sequencing matters more than the target state. Pull out the pieces that change most often first, keep them talking to the monolith through contracts you control, and make sure every increment is independently valuable. If the programme is cancelled halfway, what shipped should still have been worth doing.

In practice that has meant Play/Akka and Vert.x for reactive services, Spring Boot and Micronaut on the JVM, Docker and Kubernetes underneath, across both AWS and Azure.