Skip to content
AI Digital Hub

Infrastructure & Ops

Cloud Migration & Modernisation

Move without a weekend everyone remembers for the wrong reasons

Get a quoteBook a consultation
  • Wave-based cutover with rollback at each step
  • Cost modelled before you move, not after
  • Zero or near-zero downtime for critical paths
25-40%
Run-rate reduction post-migration

After right-sizing and reserved capacity

< 15 min
Cutover window for critical systems

Using replication and traffic shifting

0
Failed cutovers requiring emergency rollback

Because every wave is rehearsed first

Why migrations go wrong

Rarely for technical reasons. They go wrong because of an undocumented dependency, a data volume nobody measured, a vendor licence that does not permit the target platform, or a cutover plan that was written but never rehearsed.

Every one of those is discoverable in advance. The discipline is doing the boring discovery properly rather than trusting the architecture diagram, which is approximately eighteen months out of date in every organisation we have worked with.

The disposition decision

How it usually starts

  • Everything in scope, all at once
  • Sized the same as the old hardware
  • Cutover planned on a document
  • Old estate kept 'just in case'
  • Success measured by completion date

How it should run

  • Retire, retain, rehost, replatform or refactor per app
  • Sized to measured usage with committed-use pricing
  • Every wave rehearsed including rollback
  • Decommissioning is part of the plan and the budget
  • Success measured by run-rate and incident count

Waves, not a weekend

A wave is a group of applications that share data or dependencies, migrated together. Each one gets a rehearsal in a copy of the environment, a defined cutover window, explicit validation criteria, and a rollback path that stays open until the business owner signs off.

The result is undramatic, which is the point. Nobody should remember your migration.

How it runs

What the engagement looks like

Phases, not a proposal. Each one has an output you can see.

  1. 1

    Inventory and disposition

    Weeks 1-3

    Every application classified — retire, retain, rehost, replatform or refactor. A meaningful share of most estates should simply be switched off, and finding those is the cheapest win available.

  2. 2

    Build the landing zone

    Weeks 3-5

    Accounts, networking, identity, logging, guardrails and cost controls, in Terraform. Get this wrong and you spend the next two years working around it.

  3. 3

    Rehearse a low-risk wave

    Weeks 5-7

    A non-critical application first, end to end, including rollback. The purpose is to find what the plan missed while the stakes are low.

  4. 4

    Migrate in waves

    Weeks 7-20

    Grouped by coupling, sequenced by risk. Each wave has a rehearsal, a cutover window, validation criteria and a rollback that has been executed rather than described.

  5. 5

    Optimise and decommission

    Post-migration

    Right-size against real usage, apply reserved or committed-use pricing, and turn the old estate off. Migrations that skip decommissioning pay for both environments indefinitely.

FAQ

Questions we get asked

Lift-and-shift or refactor?

Usually both, applied selectively. Rehost the things that just need to be somewhere else, refactor the two or three where the cloud actually changes the economics, and retire what nobody uses. Refactoring everything is how migrations reach year three.

How do you avoid the bill going up?

By modelling it first. Cloud is more expensive than a depreciated data centre if you rehost at the same sizing and skip commitments. The savings come from right-sizing to real usage, reserved or committed-use pricing, storage lifecycle policies and switching off what you were paying for out of habit.

Can you migrate with no downtime?

For most systems, near-zero — minutes, using replication and traffic shifting. Some stateful legacy systems genuinely need a window; we will tell you which ones early rather than promising zero and negotiating later.

What about the database nobody wants to touch?

There is always one. We handle it with replication, a read-only rehearsal, validation queries agreed with the data owners, and a rollback path that stays open until you sign off. Slowly and with a lot of checking.

Talk to someone who does cloud migration

Thirty minutes with an engineer who has delivered this, not an account manager. You will get a straight answer on feasibility, rough cost and where it would fail.

Book a consultationGet a quote

Or email [email protected] · we reply within 1 business day