Home / Licensing / MSU Consumption Optimization: The Technical Levers
① Explainer · Metric and concept
The MLC bill follows one number: the peak rolling four hour average MSU. Six technical levers move that number without cutting the work the business needs. This is the map of which lever removes which cost, and the risk each one carries.
Audit notice or renewal under 18 months out? We mobilize within 48 hours.
Get expert help →Under sub-capacity Monthly License Charge, the bill for a product is set by its peak rolling four hour average MSU for the month, captured by the Sub-Capacity Reporting Tool (SCRT). That single property reframes the whole optimization problem. You are not trying to reduce total work done across the month. You are trying to reduce or flatten the peak, because the seven hundred quiet hours do not price anything. Optimization is peak management, and every lever below acts on the peak rather than on throughput.
This is also why MSU optimization is durable. The levers reduce billable MSU, not the vendor rate. Once the disciplines hold, the saving does not depend on a negotiated concession that lapses at renewal. It depends on the workload continuing to be shaped well, which is within your control.
Each lever removes a different slice of peak, with a different effort and risk profile. Sequence them by leverage and safety, not by familiarity.
| Lever | What it moves | Typical leverage | Risk if mishandled |
|---|---|---|---|
| Peak shaving | Relocates batch and reporting out of the peak hour | High | Low, if dependencies respected |
| Soft capping | Holds R4HA below a defined capacity ceiling | High | Throttling if set too tight |
| zIIP offload | Moves eligible work off the general purpose peak | Medium to high | Low, eligibility dependent |
| SCRT discipline | Keeps sub-capacity basis valid, avoids full capacity revert | Protects the whole estate | Full capacity revert if lapsed |
| Workload tuning | Cuts CPU per transaction in the peak path | Medium | Low, effort heavy |
| Capacity modeling | Times hardware and tier changes against the curve | Medium | Overspend if skipped |
Directional leverage, not a quote. Actual impact depends on how peaked the workload is and which products dominate the peak.
The decision rule that follows: start with peak shaving and SCRT discipline, because the first is high leverage at low risk and the second protects everything else from the worst case revert to full capacity. Layer soft capping and zIIP offload next. Treat tuning and capacity modeling as the slower, structural levers that compound the others over time.
Optimization decays if it is not maintained. Peak creep returns the moment a new job is scheduled into the peak window without review, repricing the month on one hour of work. Soft caps set once and never revisited either throttle work during real growth or quietly stop binding as the workload shifts, so the saving evaporates. zIIP eligible work can drift back onto general purpose processors after a software change. And a single lapsed or incomplete SCRT submission can revert a product to full capacity, undoing every other lever at once. The common thread is that these levers are operating disciplines, not one time projects.
Put a monthly cadence around the peak: review what set the rolling four hour average for each product, confirm nothing new crept into the window, and check that caps still bind where intended. Maintain a clean SCRT process with no late submissions. Keep peak shaving and zIIP offload under continuous review rather than set and forget. Model the regressive tiers, covered on our AWLC page, before any capacity or hardware change. And at renewal, test whether Tailored Fit Pricing changes which levers matter, since it prices an annual baseline rather than a monthly peak.
MSU consumption optimization is the practice of reducing the peak rolling four hour average MSU that prices sub-capacity Monthly License Charge software, without cutting the work the business needs. Because the bill follows a peak, the goal is to lower or flatten that peak through scheduling, capping, offload, tuning, and clean reporting, rather than to reduce total work done over the month.
For most estates it is peak shaving: moving non urgent batch and reporting out of the existing peak window so they never contribute to the billed hour. Because one peak hour can price the whole month for a product, relocating the work that lands in that hour often delivers the largest reduction for the least operational risk. Soft capping and zIIP offload usually follow.
Soft capping holds the rolling four hour average below a defined capacity ceiling, which constrains how much general purpose capacity a workload can draw during a sustained peak. Set well, it trims cost with no service impact. Set too tight, it can throttle work during a genuine spike. The lever is safe when the cap is modeled against real workload patterns and reviewed, and risky when it is set once and forgotten.
It depends on the starting position. Moving from full capacity to a disciplined sub-capacity posture commonly saves 30 to 50 percent. Within an already sub-capacity estate, layered peak shaving, capping, and zIIP offload typically remove a further increment that varies with how peaked the workload is. The levers act on billable MSU, not on the vendor rate, so the savings are durable once the disciplines hold.
Related metrics: Monthly License Charge, specialty engines, and Capacity on Demand. Where these products live: z/OS licensing. Put it to work: IBM cost optimization.