Cloud migration without downtime: a 6-phase playbook

“Lift and shift over a weekend” is how migrations go wrong. A calm, phased move keeps customers online and your team sane. Here’s the playbook we run on every engagement.

The six phases

1. Assess — inventory every workload, dependency and data store. Rank by risk and value.
2. Plan — choose a target architecture and a per-workload strategy (rehost, replatform, refactor).
3. Prepare — set up landing zones, networking, IAM and infrastructure-as-code.
4. Migrate — move in small, reversible waves. Start with the low-risk workloads.
5. Validate — test performance, security and cost against agreed targets.
6. Optimise — right-size, automate scaling and remove what you no longer need.

Keeping downtime at zero

The trick is incremental cutover: run old and new in parallel, route a slice of traffic to the new system, watch the metrics, then shift the rest. Database replication and feature flags make rollback a non-event rather than an emergency.

What “done” looks like

A migration isn’t finished when the servers move — it’s finished when costs are optimised, the platform is observable, and your team can operate it confidently without us in the room.