Mainframe optimization is an engineering discipline, not a discount talk
Mainframe software cost is driven by a measured capacity number in MSUs, not by users or cores, so optimization is engineering: the bill follows the busiest sustained interval, and one uncontrolled overnight batch peak can set the charge for the whole machine. Optimize the capacity profile and the charge falls; ignore it and you pay for one busy minute every month.
Prepared by Redress Compliance · August 7, 2026 · IBM advisory. Based on 20 to 30 mainframe estates reviewed 2024 to 2025.
Executive summary
Capacity engineering cut the bill 12 to 25 percent without removing any function. The peak is an engineering target: profile the intervals that set it, shift discretionary batch out of those windows, and apply soft caps where the workload allows.
In our reviews, a single uncontrolled batch peak set the monthly charge for the whole machine in 30 to 50 percent of estates, the median MSU bill cut from peak shaping ran around 20 percent, and moving one batch window often saved more than a year of discount negotiation.
Two charge models behave differently, and the lever differs with them.
Monthly License Charge software, the operating system and subsystems, bills monthly on the rolling capacity peak, so its lever is peak shaping.
International Program License Agreement products, the tools and utilities, license up front with optional sub capacity terms and ongoing support, behaving like perpetual plus maintenance, so their lever is capacity right sizing.
Knowing which products on the estate carry which behavior is the first step, because the same discipline prices differently on each side of the line.
Sub capacity reporting is where achievable savings quietly go unclaimed.
Paying on partition peaks instead of full machine capacity requires clean, timely Sub Capacity Reporting Tool output, and a lapse can revert the estate to full capacity: misconfigured reporting left 10 to 20 percent of achievable savings on the table in our reviews.
Aggregation across machines in one location can lower the effective rate further, depending on how workloads distribute.
And Tailored Fit Pricing swaps the peak model for a consumption baseline that helps spiky estates and penalizes flat ones, with the baseline measurement window setting the cost for the term.
The ISV contracts dwarf the platform, and they renew unbenchmarked.
Independent software vendor contracts on the mainframe, negotiated separately from the platform, ran 1.5 to 3 times the platform software cost and were renewed without benchmarking in our estates, roughly twice the platform line at the median.
The optimization program that stops at the IBM bill has optimized the smaller half of the budget, and the ISV renewals respond to the same capacity discipline, because most price on the same measured MSUs.
The two charge models, compared
| Dimension | MLC | IPLA |
|---|---|---|
| Charge cadence | Monthly | Up front plus support |
| Driver | The rolling capacity peak | Licensed capacity |
| Typical products | Operating system and subsystems | Tools and utilities |
| Main lever | Peak shaping | Capacity right sizing |
| Reporting dependency | SCRT for sub capacity | Optional sub capacity terms |
The rolling peak is the whole game on the MLC side. The charge uses the highest sustained capacity across a defined rolling window, so one uncontrolled spike, often an overnight batch, prices the entire month regardless of what every other interval did.
The three properties to hold in mind: peak driven, the busiest sustained interval sets the charge; reporting dependent, sub capacity needs clean tool output on time every month; and aggregation aware, grouping machines in one location can move the effective rate.
The reduction sequence, profile, shift, cap
- Profile: identify the intervals that set the peak, because the target has to be seen before it can be shaped.
- Shift: reschedule discretionary batch away from the online peak window, spreading load so no single interval spikes.
- Cap: apply defined capacity limits that hold partitions below a chosen MSU level, set with care so the bill is protected without starving critical work.
- Fix the reporting: confirm sub capacity submissions are configured and on time monthly, the most common reason achievable savings go unclaimed.
- Then, and only then, negotiate: the discount conversation opens after the capacity profile is optimized, against a bill already 12 to 25 percent smaller.
The IBM Z mainframe negotiation brief
The charge models, the capacity engineering sequence, the TFP decision, and the negotiation calendar worked on a representative estate.
Get the white paper →Tailored Fit Pricing, and the ISV half of the budget
Tailored Fit Pricing replaces the rolling peak with a consumption baseline set from a historical measurement period, charging growth above it: the peak penalty disappears, which suits spiky, peak driven estates, while predictable flat estates see little gain.
And the baseline is the risk, because a window measured during an atypical peak season locks higher cost for the term.
Model the current charge model against TFP on a low year and a high year before committing.
The ISV layer needs the same attention at larger scale: the independent vendor contracts, negotiated separately from IBM, ran 1.5 to 3 times the platform cost in our estates and renewed without benchmarking, yet most price on the same measured MSUs.
So the capacity discipline flows through to them too.
The operational levers in detail, workload shift, zIIP offload, and the ILMT discipline, sit in the MSU reduction guide; the strategy layer above them in the mainframe CIO advisory; and the largest ISV renewal of most estates in the Broadcom CA pricing analysis.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across mainframe estates, 2024 to 2025
Across roughly 20 to 30 mainframe estates we reviewed between 2024 and 2025, capacity engineering cut the monthly software bill 12 to 25 percent without removing any business function:
From peak shaping alone: batch retiming and soft caps, before any negotiation opened.
The separately negotiated vendor contracts, renewed without benchmarking at the median.
The standard line from the platform account team, that mainframe cost is essentially fixed and the only real lever is a periodic discount at renewal.
Failed in our estates in a measurable way: the schedule moved the bill more than the negotiation did, because the charge is set by the busiest sustained interval and that interval is under the customer's control.
On the mainframe you do not negotiate your way to a lower bill first, you schedule your way there, and the peak you control is the charge you avoid. The discount talk still happens, but it happens last, against an optimized profile, which is also when it goes best.
Your first five moves
- Profile the intervals that set your rolling peak, because 30 to 50 percent of estates were priced by one uncontrolled batch window.
- Shift discretionary batch off the peak and apply soft caps, the engineering that cut bills 12 to 25 percent.
- Verify sub capacity reporting monthly, where misconfiguration left 10 to 20 percent of achievable savings unclaimed.
- Model TFP against a low year and a high year before trading the peak model for a baseline.
- Benchmark the ISV renewals, the 1.5 to 3x of the budget nobody was pricing. The IBM practice runs the program with you.
Frequently asked questions
How is mainframe software licensed?
On measured capacity in MSUs, not users or cores, under two models: Monthly License Charge software, the operating system and subsystems, bills monthly on the rolling capacity peak, while International Program License Agreement products license up front with optional sub capacity terms plus support.
The lever differs by model, peak shaping for MLC and capacity right sizing for IPLA.
How do you reduce mainframe software costs?
By shaping capacity, not removing function: profile the intervals that set the rolling peak, shift discretionary batch out of those windows, and apply soft caps where workloads allow.
Capacity engineering cut monthly bills 12 to 25 percent in our reviews, with the median MSU cut around 20 percent, and the discount conversation opens after the profile is optimized, not before.
What is sub capacity pricing and what breaks it?
Paying on partition peaks rather than full machine capacity, dependent on clean, timely Sub Capacity Reporting Tool output: a reporting lapse can revert you to full capacity, and misconfigured reporting left 10 to 20 percent of achievable savings unclaimed in our estates.
Aggregating capacity across machines in one location can lower the effective rate further, depending on workload distribution.
Is IBM Tailored Fit Pricing worth it?
For spiky, peak driven estates it can be, because the consumption baseline removes the single interval penalty; flat, predictable estates see little gain, and the baseline is the risk, since a measurement window set during an atypical peak season locks higher cost for the term.
Model both charge structures on a low year and a high year before committing.
How big are mainframe ISV costs compared to the platform?
Larger: the independent software vendor contracts, negotiated separately from IBM, ran 1.5 to 3 times the platform software cost in our estates, roughly twice at the median, and they renewed without benchmarking.
Most ISV pricing follows the same measured MSUs, so peak shaping flows through to those contracts too, and an optimization program that stops at the IBM bill has covered the smaller half of the budget.
Can one batch job really set the whole month's mainframe bill?
Yes: MLC charges follow the highest sustained capacity across a rolling window, so one uncontrolled spike, typically an overnight batch stacked onto the online peak, prices the entire month, and that was the situation in 30 to 50 percent of the estates we reviewed.
Moving one batch window often saved more than a year of discount negotiation, which is why the peak is an engineering target first.
