Right-sizing and SQL Server consolidation cut a fibre operator's AWS cost by 57%
The challenge
Half the discovered estate did not need to move. Discovery identified 45 Windows servers and 71 Linux servers as out of scope, plus four machines running at zero utilisation. Migrating all 230 instances would have roughly doubled the annual bill for no operational return whatsoever, and every downstream figure in the business case would have inherited that inflation.
The remaining estate had a second issue. Storage, not compute, was the dominant cost. The operator had 332 TB provisioned against 184 TB actually used, and once compute was right-sized, storage would account for the majority of the bill.
The SQL Server footprint was badly over-licensed. Six instances carried 24 Enterprise cores and 8 Standard cores, against workloads that measured utilisation did not justify. SQL Server requires a minimum of four core licences per instance regardless of how few vCPUs it runs, so small database servers quietly consume licences they never use.
Windows Server licensing offered no relief through Dedicated Hosts. The estate had already moved on: 37 servers were running Windows Server 2022 and 9 on Windows Server 2025, both past the 2019 version lock that mobility to Dedicated Hosts requires.
Our approach
We scoped first. The estate was reduced to the 114 instances that genuinely warranted migration, with out-of-scope and zero-utilisation machines documented rather than quietly dropped.
We right-sized against measured utilisation rather than provisioned specification, and modelled the estate a second time as a direct match with no right-sizing, so the operator could see precisely what the analysis was worth.
We ran the full SQL Server optimisation sequence: right-sizing, CPU optimisation using EC2 Optimize CPUs, then consolidation of undersized database instances to eliminate wasted four-core minimums.
We priced Dedicated Hosts and advised against them. With most of the estate past Windows Server 2019, the cost of running Dedicated Hosts would have exceeded any licence saving available. Recommending the option would have produced a worse outcome and a less credible model.
The results
The recommended position sits 42% below the committed direct-match figure and 57% below direct match on demand.
Base compute fell from $287,906 on demand at direct match to $66,139 committed after right-sizing, a 77% reduction on the compute line alone.
SQL Server analysis reduced the forward requirement from 24 Enterprise cores to 4 and from 8 Standard cores to 4. That releases 20 Enterprise and 4 Standard cores against future purchases and Software Assurance renewals, and removes the $30,485 annual SQL Server licensing line from the AWS bill entirely.
The assessment also flagged eight servers running Windows Server 2012 with no vendor support and 56 on Windows Server 2016 approaching end of extended support, giving the operator a modernisation sequence tied to the migration rather than a separate project competing for budget later.
Why it matters
Storage represented $167,140 of the recommended $292,210, or 57% of the annual cost. Once compute is right-sized properly, the storage strategy becomes the business case. Identifying that during assessment rather than after migration changed where the operator focused their optimisation effort, and gave them a clear target: 332 TB provisioned against 184 TB used is a reclamation exercise worth six figures a year.
The scoping discipline mattered just as much. The cheapest workload to migrate is the one you decide not to migrate.
Collection for this engagement ran through Cloudamize. What that tool can and cannot see: read the tool page →
Talk to us about your migration scope
An AWS Optimization and Licensing Assessment measures actual utilisation across your estate, identifies what should not move, and prices the difference between lift-and-shift and a right-sized target before you commit.