Object storage
Rarely worth moving. It's the one cloud service where the price is genuinely competitive and the operational burden of replacing it is real.
Cloud to bare metal
Off the meter, without a rewrite.
Dedicated hardware is dramatically cheaper for steady, storage-heavy workloads, and considerably worse for some of what you're running. The engagement starts by working out which is which.
The engagement
Four phases, in order.
Each phase produces something you could act on without commissioning the next one. If the assessment says don't move, that's a successful engagement and we'll say so in writing.
- 01
Assessment
What you run, what it costs today, and what it would cost on dedicated hardware, with the operational headcount included, because that's where these business cases usually go wrong.
- 02
Target design
The architecture you'd actually move to: capacity, failure domains, storage, networking and how it gets operated once it's yours.
- 03
Cutover plan
Sequencing, data movement, rollback points and the order things move in, written so your team can execute it rather than so we're needed for every step.
- 04
Run-in
We stay through the first weeks, when the surprises turn up. Handover happens after it's stable, not on the day it's technically live.
The honest part
What we'll tell you not to move.
A wholesale exit is usually the wrong answer, and anyone selling you one has an incentive we don't. Most successful migrations are partial: the steady, predictable, storage-heavy workloads move, and the rest stays where elasticity is worth paying for.
Anything with spiky, unpredictable demand
Dedicated capacity is cheaper when it's busy and wasted when it isn't. Bursty workloads are exactly what elastic infrastructure is good at.
Managed services you'd have to rebuild
If replacing a hosted service means running a database team you don't have, the saving is a transfer of cost rather than a reduction in it.
Where the savings actually come from
Steady load, heavy storage.
The workloads that pay for themselves fastest are the ones running at a consistent level with a lot of data attached: log stores, metrics, search indices, analytics. They're billed by the gigabyte in the cloud, and by the machine on dedicated hardware.
It's also why we run our own platform this way, which is a reasonable thing to ask us about on the call.
Counted properly
A migration business case that ignores the operational headcount isn't a business case. We include it, and occasionally that's the number that says don't do it.
Find out whether it's worth it.
The assessment is fixed-scope and stands on its own. You get a defensible number either way, and no obligation to act on it.