On-premises to AWS migration

A migration from on-premises to AWS, using zero-downtime deployment and cost optimization strategies.

Hardware refresh cycles, capacity limits and slow provisioning are the usual drivers. The risk is treating the move as one event instead of a controlled series of reversible steps.

What a zero-downtime migration requires.

• –A rollback path at every step, until the old system is retired
• –Continuous data replication with measured lag
• –A dependency map, so nothing is moved before what it needs

Source and target run in parallel, connected by private networking, with data replicating continuously. Weighted DNS moves traffic gradually, so reverting a step is just another weight change.

The strategy is chosen per application, not for the whole estate.

• –Rehost when speed matters and the application is stable.
• –Replatform to managed databases where the operational saving is clear.
• –Retire whatever discovery shows nobody uses.

Discover, then group applications into waves by dependency. Move a low-risk wave first to prove the runbook, then improve it after every wave.

Each application is cut over using a rehearsed runbook: reduce DNS TTLs in advance, confirm replication lag, shift traffic in steps, and watch the same SLOs on both sides.

Private connectivity through VPN or Direct Connect, encryption in transit and at rest, and access redesigned around IAM instead of copied firewall rules.

Baseline performance before moving, so "slower than before" is measurable. Keep the source system running through a soak period after each cutover.

Typical surprises, and how they are managed.

• –Undocumented dependencies: confirm them from network flow data rather than from documents.
• –Chatty applications split across the link: move tightly coupled systems in the same wave.
• –Costs after a lift-and-shift: rightsize from real data after the move, not before.

No verified figures are published for this project. The numbers that would prove it worked: applications moved, downtime per cutover, and run-cost before and after.