AWS Windows
Infra Optimization
The Qlik environment ran at full capacity all day and all night, whether or not anyone was using it. By scheduling non-production compute, optimizing hyper-threading and rightsizing servers against real usage data, we brought annual infrastructure spend from about $1.08M to about $640.5K. That's roughly $439K saved every year, a 40% reduction. The client didn't need a budget to get started, and our team carried out the changes.
At a glance
-
Annual infrastructure spend $1.08M → $640.5K (before → after)
-
Annualized savings $439K
-
Reduction 40%
-
Waste identified in the first audit 32%
-
Time to savings 60 days
-
Main levers
- Scheduled start/stop
- Hyper-threading optimization
- Rightsizing
Where The Money Was Going
The Qlik infrastructure was set up to run continuously, so capacity was billed even when workloads didn't need it. This was most visible in non-production environments, which stayed on during long stretches when no one was using them. Because the servers ran on Windows, the extra capacity also kept licensing costs higher than the workload required.
Nothing was broken. The environment was simply set up for constant capacity rather than actual demand, and the cost reflected that.
How the engagement worked
The project ran on our Save Now Pay Later model. The client paid nothing upfront and took on no financial risk. We handled the audit and the implementation ourselves, and our fee came only from savings that had already shown up on the cloud bill. If the bill hadn't come down, there would have been no invoice.
Our engineers made the changes alongside the client's team, so their engineers only needed to put in a small amount of time. Many Optimization tools are good at showing where money is wasted, but they stop there. Our job was to make the changes and keep the savings in place.
Always On
(High cost)- Servers run 24/7Even when there's no activity
- Pays for unused capacityIncludes non-production environments
- Higher licensing costsDue to extra Windows capacity
Scheduled & Optimized
(Real usage)- Runs on actual demandSchedule to shut down during inactive hours
- Reduces unnecessary costsLess infrastructure and licensing spend
- Optimized capacityScales when needed, not 24/7
Save Now Pay Later
- No upfront cost
- No financial risk
- Fee only from realized savings
If there are no savings, there's no invoice.
How we approached it
We didn't make broad cuts. We started with the data and changed only what the usage numbers supported.
- 1
Utilization analysis
We reviewed how each environment was used across the day and week to see where capacity sat idle.
- 2
Scheduled start and stop
Non-production capacity now runs when teams need it and shuts down when they don't.
- 3
Hyper-threading optimization
We adjusted hyper-threading settings to reduce compute consumption and bring Windows licensing costs under control.
- 4
Rightsizing
We resized the remaining servers to match the workloads they actually carry.
What changed for the team
Compute and licensing costs now follow actual workload requirements. The team runs the environment on a schedule it controls, with clear visibility into what each part of the infrastructure costs and why. The savings our fee doesn't take stay with the client to put toward other work.
Smarter infrastructure.
Real savings.