The Server You Might Not Need to Buy
Before you sign off a hardware refresh or another round of Enterprise licences, it is worth knowing whether the workload actually requires them.
Why On-Premise Waste Is Harder to See
In Azure, inefficiency is metered and billed every month, so somebody eventually asks about it. On your own hardware it hides.
A query reading ten times more data than it needs to costs nothing visible when the server has the headroom to absorb it. Nobody is invoiced for a scan. The cost only appears years later, as a capital request: the system is slowing down, the server is at end of life, and here is a quote for the replacement.
By that point the specification has usually been sized against current behaviour, on the reasonable assumption that current behaviour is necessary. Frequently it isn't. The new server is being sized to run the same inefficiency faster, and in three years the same conversation happens again with a larger number attached.
Which is why it's worth asking one question before the purchase order goes out: how much of this capacity is for the business, and how much is to cover work the database shouldn't be doing?
Where the Money Actually Goes
Cores You Pay For Twice
SQL Server is licensed per core, so every core in that specification is billed once by the hardware supplier and again by Microsoft. Oversizing the box is not a one-off cost.
Enterprise You May Not Need
Enterprise edition is a substantial multiple of Standard. Sometimes the workload genuinely requires it. Often it is there because of one feature, or because nobody has checked since the last renewal.
Memory Buying Time
More RAM makes a badly indexed workload tolerable by keeping more of it in cache. It treats the symptom convincingly enough that the underlying cause is never investigated.
Storage Full of History
Years of data nobody queries, sitting on your most expensive tier, inflating the size of every backup, every restore test, and every maintenance window.
Indexes Nobody Uses
Every insert and update maintains every index, including the ones no query has touched in years. They cost write performance, storage, and backup time, in perpetuity.
Instances That Should Be One
Servers bought one project at a time, each sized for its peak, each licensed separately, most of them idle most of the day.
What We Actually Do About It
We look at what the workload is really asking of the machine: what is being read, how often, and why. Then we work out how much of that is avoidable. Not a spreadsheet exercise, and not a licensing trick.
Sometimes the answer is that the estate is running well and the hardware is genuinely at its limit, in which case you buy the server knowing it is the right decision rather than hoping. More often a proportion of the demand turns out to be avoidable, and the question changes from which server to buy to whether one is needed yet.
Either outcome is useful. The expensive version is the one where nobody asked.
Where a move is on the cards anyway, the same analysis feeds directly into planning the migration or upgrade, since the worst thing you can do is carry the inefficiency across and size the new platform around it.
Find Out Before You Commit
Our free health check looks at one instance and comes back with a plain-English report on what is healthy, what is quietly costing you, and what it would take to put right. If you are holding a hardware quote, that is a sensible thing to read first.
For a whole estate, the SQL Server Audit covers everything end to end with a written, prioritised plan and a fixed fee agreed up front. Against a six-figure refresh, it is a rounding error that occasionally removes the need for the refresh altogether.
Holding a Quote You're Not Sure About?
Send us the specification and what it is meant to fix. We'll tell you honestly whether we think you need it.
Talk to UsOr call us directly on 01293 988004.