Data-Center Exit to AWS

01 / The Challenge

A data-center lease expiring in nine months put a hard deadline on Atlas's infrastructure. Point-of-sale, e-commerce, and inventory systems for 46 stores had to move without disrupting trading and the CFO wanted the cloud bill lower than the lease, not higher.

02 / Our Approach

The thinking behind the build

Hard-deadline migrations fail on surprises, so we spent the first six weeks purely on discovery: every workload inventoried, every dependency mapped, every cutover rehearsed in a staging environment before it happened for real. Wave order was chosen by risk boring internal systems first to prove the pipeline, revenue-critical retail systems last, once the process was routine.

Cost discipline was engineered in from the start rather than optimized later: right-sizing, reserved capacity, and per-business-unit cost tagging were part of each workload's migration checklist, not a post-launch project.

03 / The Solution

What we built

We stood up a secure AWS landing zone as code, mapped dependencies across all 70 workloads, and executed a five-wave migration with rehearsals and rollback plans for each. Right-sizing, reserved capacity, and storage tiering were engineered in from the start, with per-business-unit cost visibility.