Moving from legacy to cloud: a practical playbook
Legacy systems are the silent cost centre most companies don't measure: slow releases, expensive servers, and the one engineer who knows how it all fits together. Moving to the cloud feels like a huge project — but it doesn't have to be a big-bang one.
Lift first, optimise later
The cheapest, lowest-risk first step is almost always a straight lift-and-shift: move the existing system onto cloud infrastructure with as little change as possible. You gain reliability, monitoring and cost visibility immediately.
Only once it's running in the cloud do we start modernising — and we do that one component at a time, so there's always a working system in production.
Strangler-fig, not rewrite
A full rewrite is how legacy migrations die. Instead, we use the strangler-fig pattern: build the new capability alongside the old system, route a little traffic to it, prove it works, and gradually retire the old code.
Every step is reversible. If something goes wrong, you're never more than one rollback away from a system that still works.
Measure the win
A migration isn't done when the last server is switched off; it's done when the numbers say it was worth it. We set three baselines before we start — deploy time, incident count, and monthly infrastructure cost — and track them monthly.
Typically, we see deploy time drop from hours to minutes, incidents fall sharply once monitoring exists, and cost becomes predictable instead of a monthly surprise.
Written by the Sadhna Infosys team.