The Optimization and Licensing Assessment
A complete picture of what you run, what it uses, what it would cost on AWS, and what you have already paid Microsoft for. Funded by AWS, delivered by us, and yours to keep whatever you decide.
The question an OLA exists to settle.
Should this estate move, what should it look like when it does, and what will it actually cost to run?
Almost every migration business case gets built on two assumptions that nobody checks. That the new environment should be sized like the old one, and that every Windows and SQL Server instance needs a licence bought from the cloud provider. Both are usually wrong, and both are wrong in the same direction.
An OLA replaces both assumptions with measurement. It is the only point in a migration where correcting them is free.
Five phases, in this order, for a reason.
Scope
Before anything is sized we agree what is genuinely in scope. Decommissioning candidates, servers already under shutdown review, duplicate records and conflicting power states come out first, because every one of them inflates the figures downstream. On one engagement this removed half the discovered estate.
Discover
Inventory across on-premises, VMware, Hyper-V and existing cloud. We work from whatever collection method suits your change-control posture, including data you have already collected with someone else's tooling.
Measure
Utilisation captured over a representative period rather than read from a specification. This is the phase that takes calendar time, and it is the phase that makes the rest of the model defensible.
Model
Right-sizing against measured load, then a full AWS cost model. We price several commercial positions against one another rather than presenting a single number, so you can see what each individual decision is worth.
License
Your Microsoft entitlement read against the modelled estate. Which licences are eligible to move, what tenancy model they require, where edition or core-minimum optimisation changes the answer, and what all of that is worth against licence-included pricing.
The part most assessments treat as an appendix.
A migration model that ignores your entitlement will quietly have you rent software you have already bought, every hour, for as long as the workload runs.
Entitlement mapping
What you own, read from the licence statement rather than assumed from the estate. Windows Server Standard and Datacenter, SQL Server Standard and Enterprise, core counts and Software Assurance status.
Mobility eligibility
Which licences can move to AWS and under what conditions. Licence mobility through Software Assurance, and the separate rule that Windows Server licences must predate October 2019 to run on Dedicated Hosts.
Tenancy modelling
Shared tenancy, Dedicated Hosts, and mixed designs where a subset of the estate is packed onto hosts under BYOL while the remainder stays shared. The mixed model is frequently the cheapest and almost never the default.
Edition and core optimisation
SQL Server Enterprise instances with no technical dependency on Enterprise features, four-core minimums, and consolidation that releases licence cores against future true-ups and renewals.
A migration business case
Several commercial positions priced against one another, with the baseline for each stated.
A right-sizing schedule
Instance, storage and database recommendations per workload, against measured utilisation rather than provisioned specification.
A licensing position
What you own, what is eligible to move, the tenancy model that makes it work.
A scope recommendation
The workloads we would not migrate, and why. Retirement, consolidation and leave-in-place candidates, named individually.
A walkthrough
A working session on the model itself rather than a document sent by email. Every assumption is open to challenge, and several usually are.
Less than you think, and probably some of it already exists.
Collection is read-only. Agentless options exist for every major platform, and in regulated estates there are routes that require no cloud connectivity at all.
If you ran a discovery exercise in the last two years, that data is very likely still usable, even where the tooling behind it has since been retired.
The questions we get asked first.
What if the answer is that we should not migrate?
Then that is the answer, and we will write it down. Several of our published engagements recommended leaving substantial parts of an estate exactly where it was.
We already ran an assessment. Is this a repeat of that?
Usually not. Most assessments size the estate and stop. The licensing analysis is a separate discipline and it is frequently missing entirely, which is where the largest single saving tends to sit. If your existing data is sound we will work from it rather than collecting again.
Start with the data you already have.
Tell us what you are sitting on and we will tell you what it can and cannot support before anything is committed.