Plenty of data, no story
The most common mistake in Azure adoption? Treating lift-and-shift as the final destination instead of a stepping stone. Teams that copy their on-premises specs straight over to Azure VMs structurally end up buying forty to sixty percent more capacity than they need. Not out of carelessness, but because the on-premises mindset comes along for the ride. Henk van der Valk, co-founder of OptimaSure, sees this pattern in nearly every migration. In this blog, he explains the three cost killers that sneak in, and how to deal with them.
You’ve made the call. The on-premises SQL Servers are moving to Azure. The business case is signed off, Azure Migrate is ready to go, and within a few months the workloads are running in the cloud. Mission accomplished.
Three months later, the first real invoice lands, and it’s considerably higher than anyone expected.
On-premises SQL Servers are almost always generously sized. Makes sense: you buy hardware for three to five years, so you size for the peak you might see two years from now. Plus a margin. Plus some extra room for that year-end close in December. Plus a buffer, just in case.
In your own data center, that’s not a problem. The hardware is already depreciated, and your power bill barely changes whether your server is running at 20% or 70% CPU. Idle capacity is free. In Azure, idle capacity is a fixed monthly invoice.
What Henk sees in practice: teams that copy their on-premises specs over to Azure end up buying too much, because the mindset never shifts. On-premises, you size for the peak. In Azure, you size for the pattern.
The wrong VM family.
This is the quiet killer. SQL Server workloads are memory-bound in ninety percent of cases, not compute-bound. And yet databases keep landing, time and again, on general-purpose dev/test workhorse VMs with plenty of cores but relatively little memory per core.
The result: SQL Server can’t cache enough data, has to read from disk more often, queries get slower, disk burst capacity isn’t available, and the natural reaction is “this VM is too small, scale it up.” Now you’re paying for even more cores you don’t use, while the actual problem was memory. There are dozens of SKUs to choose from instead (the B-series and E-series, for example) — memory-optimized, higher IOPS limits, and often cheaper than a scaled-up D-series trying to reach the same amount of memory through brute force.
Sizing for the peak instead of the pattern.
Your on-premises SQL Server peaks at 75% CPU for half an hour a day. The rest of the time, it idles around 8%. In the cloud, you pay for that 75% capacity, 24 hours a day, seven days a week.
That’s not a shortcoming of Azure, it’s a missed opportunity at migration time. The cloud is built around the idea that you scale up and down. If you don’t, you’re paying premium prices for on-premises behavior. For variable workloads, B-series burstable VMs exist, which build up credits during quiet hours and spend them during peaks. For dev, test, QA, and batch environments, Azure SQL Database Serverless delivers an immediate win: it automatically scales back to zero after office hours, when your people are off, and you pay significantly less.
Storage that doesn’t match the workload.
In your own data center, storage was set up once and never touched again. In Azure, storage is an active choice with direct cost consequences: Standard SSD disks capped at 500 IOPS versus Premium SSDs at 5000, Ultra Disk, host caching on or off, data and log files on separate disks. Accepting the default settings almost always means paying too much, running queries too slowly, or both at once.
This is the core of it. Lift-and-shift is technically a migration, but economically it’s a shift. On-premises, you bought capacity (CapEx), depreciated it, and used it as much as possible to earn back the investment. In Azure, you consume capacity (OpEx), which means you need to do the opposite: use as little as possible for as much business value as possible.
That calls for different reflexes. Size for the pattern instead of the peak. Set up autoscaling instead of provisioning “just in case.” Switch workloads off wherever you can, instead of leaving them on because you can. Evaluate the architecture continuously instead of once every three years.
These are reflexes on-premises teams never had to develop, because it never mattered. In Azure, it matters every single month. Literally, on the invoice.
Let’s clear up a misunderstanding: lift-and-shift isn’t a bad strategy. It’s often the fastest, safest way to get out of an aging data center, to escape an expiring Windows Server or SQL Server support contract, or to leave VMware licensing behind. For many organizations, it’s the only realistic first step.
The mistake is treating lift-and-shift as set in stone. If you don’t start right-sizing, monitoring, and optimizing immediately after the migration, you end up, in practice, paying premium cloud prices for on-premises overcapacity for the next three to five years. That’s an approach problem, not an Azure problem.
A well-run Azure environment for data workloads needs three things. Ongoing insight into what you’re using versus what you actually need. VM choices that match the real character of your workloads. And treating review and right-sizing as a standing responsibility, along with the nerve to actually act on those insights: scaling back, switching off, re-architecting where it pays off.
At OptimaSure, Henk brings all of that together in one approach. The Azure Cost Scan maps out within 30 days exactly where your money is leaking: oversized VMs, the wrong hardware family, idle resources, storage hotspots. CloudXcellence, OptimaSure’s own tooling, connects to your Azure environment within a day and shows you what you can save, by resource category and by individual resource. You click straight through to action. No ticket, no queue, no report sitting in a drawer.
For the data-specific side (SQL Server configuration, indexing strategy, query optimization) OptimaSure brings in the DBA expertise of sister company OptimaData. Because without healthy databases, there’s no healthy cloud invoice.
Lift-and-shift is the fastest way to get out of your data center. Without a follow-up step, it’s also the most expensive way to get stuck in Azure and that bill keeps running, month after month.
Saving twenty to thirty percent on your Azure invoice isn’t an optimistic assumption for most organizations. It’s low-hanging fruit. The only question is: who’s going to pick it?
Want to know where your Azure environment is leaking? Request an Azure Cost Scan. Insight within 30 days, structural savings within 31.