How Redress Compliance helped Mizuho Financial Group — one of Japan's largest financial institutions — achieve $71 million in verified IBM z/OS mainframe licensing and support savings through MLC optimisation, workload realignment, LPAR reconfiguration, and strategic contract restructuring.
📖 This case study is part of our comprehensive IBM Mainframe Software Licensing — A CIO's Advisory guide — covering MLC pricing mechanics, workload management, sub-capacity optimisation, and contract negotiation strategies for enterprises running IBM z/OS mainframe environments.
Mizuho Financial Group is one of Japan's three "mega-bank" financial groups and one of the largest financial institutions in the world. Headquartered in Tokyo, Mizuho operates across corporate banking, retail banking, investment banking, asset management, and securities — serving millions of customers through a global network spanning Asia, the Americas, and Europe. With over 60,000 employees and total assets exceeding ¥250 trillion (approximately $1.7 trillion USD), Mizuho is a systemically important financial institution (SIFI) subject to the most stringent regulatory requirements in the global banking sector.
For decades, IBM mainframe infrastructure has served as the critical foundation for Mizuho's core banking operations. The bank's z/OS-based systems process millions of transactions daily — including real-time payment processing, securities settlement, foreign exchange operations, loan origination, regulatory reporting, and customer account management. Mizuho licences a substantial portfolio of IBM z/OS software — including DB2 for z/OS, CICS Transaction Server, IMS (Information Management System), MQ for z/OS, and numerous proprietary IBM middleware and utility products — running on high-capacity IBM zSeries systems across global data centres.
By 2023, Mizuho's combined IBM mainframe licensing and support costs had reached levels that the bank's technology leadership considered unsustainable — representing a disproportionate share of the total IT budget relative to the strategic flexibility they provided. The costs were driven by IBM's Monthly License Charge (MLC) pricing model, which ties software costs to peak processor utilisation, combined with long-standing contract structures that included bundled products, limited termination flexibility, and pricing terms that had not been renegotiated to reflect current market conditions.
Mizuho engaged Redress Compliance to conduct a comprehensive mainframe licensing and contract optimisation programme. Over a sixteen-week engagement, Redress analysed every element of Mizuho's IBM mainframe licensing, identified systematic inefficiencies in workload management, MLC measurement, LPAR configuration, and contract structure, and developed and implemented a multi-track optimisation strategy. The result: $71 million in verified, defensible, and recurring savings — achieved without any disruption to Mizuho's mission-critical banking operations.
Batch job scheduling realigned to reduce peak rolling 4-hour average utilisation — the metric that directly determines IBM mainframe software costs.
Logical partitions restructured to optimise sub-capacity licensing, isolating high-cost products to dedicated LPARs with controlled capacity.
Legacy bundled agreements decomposed, dormant products decommissioned, and renewal terms renegotiated based on the optimised usage baseline.
All optimisations implemented without any disruption to real-time transaction processing, settlement operations, or regulatory reporting.
Mizuho's mainframe environment is one of the largest and most complex in the global financial services sector. The bank operates multiple IBM zSeries mainframe complexes across its primary data centres in Japan, with additional capacity in the United States and Europe supporting regional operations. These systems are not legacy infrastructure in the pejorative sense — they are actively developed, continuously enhanced, and mission-critical platforms that process the real-time transactional workloads that constitute the core of Mizuho's banking operations.
Core Banking and Transaction Processing: Mizuho's mainframe systems process millions of financial transactions daily — including domestic and international payments, securities trades and settlements, foreign exchange conversions, loan disbursements and repayments, and interbank clearing operations. These workloads run on CICS Transaction Server and IMS, with DB2 for z/OS providing the database layer. Transaction volumes peak during specific windows — Tokyo market open, overlapping trading hours between Asian and European markets, and end-of-day batch settlement windows — creating the utilisation patterns that drive IBM's MLC pricing.
Batch Processing: In addition to real-time transaction processing, Mizuho's mainframes execute extensive batch processing workloads — including end-of-day settlement runs, regulatory reporting calculations (Basel III/IV capital adequacy, liquidity coverage ratios, JFSA reporting), interest calculations, statement generation, and data warehouse feeds. These batch jobs consume significant processor capacity and, critically, their scheduling relative to peak transaction processing periods has a direct impact on MLC costs.
IBM Software Portfolio: Mizuho's IBM mainframe software portfolio includes dozens of products licensed under both MLC and IPLA (International Program License Agreement) pricing models. The highest-cost products — DB2, CICS, IMS, and MQ — account for the majority of licensing spend, but the portfolio also includes IBM utilities (DFSORT, DFSMS), monitoring tools (OMEGAMON, Tivoli), and specialised banking middleware. Over the years, the portfolio had accumulated products that were no longer in active use but remained under licence and support.
¥250T+ total assets (~$1.7T USD). 60,000+ employees. Global mega-bank. SIFI-designated. Corporate, retail, investment banking, and asset management.
Multiple IBM zSeries complexes. DB2, CICS, IMS, MQ, utilities, and monitoring tools. Global data centres (Japan, US, Europe). Millions of daily transactions.
Mainframe licensing and support costs had reached unsustainable levels. MLC pricing driven by peak utilisation patterns. Legacy contracts with bundled products and rigid terms.
IBM's Monthly License Charge (MLC) model is the dominant pricing mechanism for z/OS mainframe software. MLC costs are calculated based on the rolling 4-hour average (R4HA) of peak processor utilisation during each month — specifically, the highest 4-hour average of MSU (Million Service Units) consumption within the measurement period. This means that a single 4-hour window of elevated utilisation can determine an entire month's software licensing cost — even if the system is lightly loaded for the remaining 716+ hours of that month.
This pricing model creates significant optimisation opportunities for banks with complex workload patterns. If batch processing jobs coincide with peak transaction processing — pushing utilisation into a higher MSU tier during the measurement window — the result is inflated MLC costs that bear no relationship to the average utilisation of the system. By carefully managing when workloads execute (not just whether they execute), enterprises can reduce their peak R4HA without reducing functionality — and the MLC cost reductions follow directly.
IBM's MLC pricing also uses a tiered MSU structure where the per-MSU cost decreases at higher tiers — but the absolute cost still increases. This means that reducing peak MSU consumption by even a modest percentage can produce savings that are disproportionately large relative to the capacity reduction, because the reduction eliminates the highest-cost incremental MSUs at the top of the tier structure.
⚠️ IBM MLC: The Rolling 4-Hour Average Trap: IBM's MLC pricing is determined by the highest rolling 4-hour average of MSU consumption in each month. This means a single period of coinciding batch and online workloads can inflate an entire month's software bill. Most large mainframe customers are overpaying because their workload scheduling was never designed with MLC optimisation in mind. Careful workload management — without any reduction in processing capacity — is the single highest-impact optimisation available.
📌 Related Guide: For a comprehensive understanding of IBM mainframe licensing mechanics, see our IBM Mainframe Software Licensing — A CIO's Advisory guide.
Mizuho's mainframe workload scheduling had evolved over decades without any deliberate alignment to IBM's MLC pricing mechanics. Batch processing jobs — including end-of-day settlement runs, regulatory calculations, and data warehouse extracts — were scheduled based on operational convenience and processing deadlines, not on their impact on the rolling 4-hour average. During several windows each month, batch workloads coincided with peak online transaction processing, pushing the R4HA to levels that inflated MLC costs by tens of millions of dollars annually. The bank's operations team had never analysed the relationship between batch scheduling and licensing costs — because batch scheduling was treated as an operational discipline, while licensing was a procurement function. The two teams had never been connected.
Mizuho's Logical Partition (LPAR) structure on its zSeries mainframes had grown organically over many years. High-cost IBM software products (DB2, CICS, IMS) were co-located in LPARs alongside lower-cost or non-MLC products, without consideration of the licensing implications. IBM's sub-capacity pricing determines MLC costs based on the capacity allocated to each LPAR — meaning that an LPAR running a high-cost product like DB2 alongside a low-cost utility is licensed for the combined capacity of both workloads, even though only the DB2 component carries the elevated MLC rate. Restructuring LPARs to isolate high-cost products into dedicated, right-sized partitions is a standard optimisation technique — but one that requires specialist IBM licensing knowledge to implement correctly.
Over decades of mainframe evolution, Mizuho had accumulated IBM software products that were no longer in active use but remained under licence and support. Some had been replaced by newer IBM products or third-party alternatives but were never formally decommissioned. Others were installed as part of bundled agreements and had never been deployed for any productive purpose. These dormant products generated ongoing MLC and support costs despite providing no business value. Identifying which products were genuinely dormant (as opposed to being used by a single application or process that might not be immediately visible) required careful analysis of actual product usage data.
Mizuho's IBM mainframe agreements had been structured over multiple contract generations — with bundled product packages, multi-year commitments, and pricing terms that reflected historical negotiation positions rather than current market conditions. IBM's commercial team consistently resisted restructuring proposals, instead offering "solutions" that involved migrating workloads to IBM Cloud or committing to new bundled licence renewals — approaches that would have increased IBM's recurring revenue rather than addressing Mizuho's root cost challenges. The bank needed an independent advisor who could navigate IBM's commercial resistance and negotiate contract changes based on the optimised usage baseline.
Redress began with a deep-dive analysis of Mizuho's entire IBM mainframe licensing position and workload profile. The assessment mapped every IBM entitlement across the z/OS estate — DB2, CICS, IMS, MQ, and all MLC and IPLA products — cross-referencing entitlements against actual deployments, configurations, and usage patterns. Simultaneously, Redress analysed Mizuho's peak rolling 4-hour average (R4HA) utilisation data across multiple months, identifying exactly when peaks occurred, which workloads (batch vs. online) contributed to those peaks, and how much each workload contributed to the total MSU consumption during peak windows. This analysis also reviewed IBM SCRT (Sub-Capacity Reporting Tool) data and workload management configurations. The assessment revealed that a significant proportion of Mizuho's mainframe software was over-licensed, misaligned with actual utilisation, or priced inefficiently based on configuration and workload timing decisions that had never been optimised for licensing impact.
Based on the R4HA analysis, Redress developed a workload scheduling optimisation plan designed to reduce peak utilisation without affecting processing capacity or operational deadlines. The primary optimisation targeted batch processing jobs that were contributing to peak R4HA windows. By shifting specific batch runs — particularly end-of-day settlement calculations, regulatory reporting extracts, and data warehouse loads — to time windows that did not coincide with peak online transaction processing, Mizuho could materially reduce its monthly R4HA without reducing the total amount of processing performed. Redress modelled the impact of multiple scheduling scenarios, quantifying the expected MLC reduction for each and identifying the optimal configuration that balanced licensing cost reduction with operational requirements such as processing deadlines and downstream data dependencies. The workload realignment was designed in collaboration with Mizuho's mainframe operations team to ensure that every scheduling change was operationally validated and risk-assessed before implementation.
Redress redesigned Mizuho's LPAR structure to optimise sub-capacity licensing for high-cost MLC products. The reconfiguration isolated DB2, CICS, and IMS workloads into dedicated LPARs with capacity allocations precisely matched to actual processing requirements — eliminating the licensing overhead caused by co-location with lower-cost workloads. Redress also identified opportunities to adjust LPAR capacity capping configurations, ensuring that peak MSU allocations reflected operational needs rather than historical defaults that had never been reviewed. The LPAR reconfiguration was implemented in phases — with each change validated against Mizuho's performance requirements and regulatory processing deadlines before proceeding to the next phase.
Redress identified IBM software products across Mizuho's mainframe estate that were licensed and under support but no longer in productive use. This included utility products that had been superseded by newer tools, monitoring components that had been replaced by third-party alternatives, and middleware products that had been installed as part of bundled agreements but never deployed for any business function. For each dormant product, Redress confirmed the product's inactivity through usage analysis, validated that no application dependencies existed, and documented the decommissioning recommendation with supporting evidence. The dormant product cleanup removed ongoing MLC and support costs for software that was generating zero business value — a straightforward cost reduction with no operational risk.
With the optimised licensing baseline established — reflecting the reduced R4HA from workload realignment, the improved LPAR configurations, and the decommissioned dormant products — Redress supported Mizuho's procurement and legal teams in restructuring the IBM mainframe contract. The negotiation strategy leveraged the optimised baseline as primary evidence: Mizuho's demonstrably reduced IBM consumption created a strong commercial basis for renegotiating pricing terms, removing bundled products that were no longer licensed, and securing more flexible contract structures. Redress built the commercial negotiation plan, prepared counter-proposals to IBM's standard renewal offers, and advised on when to concede and when to hold firm. The renegotiated contract reflected Mizuho's actual optimised consumption rather than the historical, inflated baseline that had governed previous agreements.
| Optimisation Strategy | Verified Savings | Method |
|---|---|---|
| MLC Peak / R4HA Reduction | ~$32M | Batch workload scheduling realigned to avoid coincidence with peak online processing; R4HA reduced materially |
| LPAR Reconfiguration | ~$18M | High-cost products isolated to dedicated, right-sized LPARs; sub-capacity licensing optimised |
| Contract Restructuring | ~$12M | Renewal pricing renegotiated based on optimised baseline; bundled products removed; improved terms secured |
| Dormant Product Decommissioning | ~$9M | Unused IBM software products identified, validated, and decommissioned — removing ongoing MLC and support costs |
| Total Verified Savings | $71M | Recurring savings — zero operational disruption |
The MLC peak reduction ($32M) represents the single largest savings category — confirming that workload scheduling optimisation is the most impactful lever available for IBM mainframe cost reduction. The savings are recurring: because MLC costs are calculated monthly based on the R4HA, the reduced peak utilisation produces savings in every subsequent month as long as the optimised workload scheduling is maintained. Mizuho's operations team was trained on the ongoing management of workload scheduling within the MLC-optimised framework, ensuring that the savings persist beyond the engagement period.
The LPAR reconfiguration ($18M) and dormant product decommissioning ($9M) represent structural optimisations that produce permanent cost reductions. Once high-cost products are isolated in correctly sized LPARs and dormant products are decommissioned, the licensing position is fundamentally improved — these savings do not require ongoing management to sustain. The contract restructuring ($12M) reflects improved commercial terms that will govern Mizuho's IBM relationship through the next contract period and beyond.
The combined optimisation programme produced $71 million in verified savings across MLC reduction, LPAR reconfiguration, dormant product decommissioning, and contract restructuring. These savings are not one-time adjustments — the MLC reductions recur monthly, the LPAR and decommissioning savings are permanent structural improvements, and the contract restructuring establishes improved pricing for the next agreement period. The $71 million represents one of the largest verified IBM mainframe optimisation outcomes for a single financial institution globally — and it was achieved entirely through operational and commercial optimisation, without any reduction in Mizuho's processing capacity, application functionality, or service levels. For a bank where mainframe software licensing represented one of the largest individual line items in the technology budget, the savings had a material impact on the overall IT cost structure and freed budget for strategic investment in digital transformation initiatives.
All optimisations were implemented without any disruption to Mizuho's real-time transaction processing, settlement operations, regulatory reporting, or any other operational function. The workload realignment changed when batch jobs execute — not whether they execute or how much processing capacity is available. The LPAR reconfiguration changed how capacity is allocated among partitions — not the total capacity available. Mizuho's core banking systems continued to process millions of transactions daily throughout the engagement. For a SIFI-designated institution where operational resilience is a regulatory requirement, this zero-disruption outcome was a non-negotiable prerequisite.
Redress transferred the knowledge and tools necessary for Mizuho's technology team to maintain the optimised licensing position on an ongoing basis. This included training on MLC peak monitoring and workload scheduling management, documentation of the LPAR configuration rationale and sub-capacity licensing implications, a product usage review process for identifying future dormant products, and guidelines for evaluating the licensing impact of infrastructure changes (new LPARs, capacity upgrades, workload migrations) before they are implemented. The framework also includes a quarterly MLC review process — a structured analysis of each quarter's R4HA data to identify any drift in peak utilisation patterns, workload scheduling changes that may have inadvertently increased peaks, or new batch jobs that have been introduced without MLC impact assessment. This sustainable framework ensures that the $71 million in savings is not eroded over time by workload drift, configuration changes, or new product deployments that are implemented without licensing awareness. In large mainframe environments, it is remarkably easy for optimisation gains to be undone incrementally — a new batch job scheduled during a peak window here, an LPAR capacity increase there — unless the organisation embeds licensing awareness into its standard operational processes.
The engagement fundamentally changed the commercial dynamic between Mizuho and IBM. Where IBM had previously driven renewal discussions based on historical (inflated) consumption baselines, Mizuho now negotiates from an optimised position with complete visibility into its actual licensing requirements. IBM's proposals to migrate workloads to cloud services or to commit to bundled renewals are evaluated against the optimised on-premises baseline — providing Mizuho with a genuine cost comparison rather than a comparison against an artificially inflated starting point. The recalibrated relationship also means that Mizuho approaches future IBM interactions — whether renewal negotiations, audit enquiries, or capacity planning discussions — with the analytical framework and commercial confidence to evaluate IBM's proposals independently, rather than relying on IBM's own assessments of what the bank needs and what it should pay.
In Mizuho's case — and in the majority of large-scale IBM mainframe environments — the rolling 4-hour average peak drives the largest portion of software costs. Optimising when workloads execute (particularly batch processing relative to online transaction peaks) produces savings that are disproportionately large relative to the operational change required. This was precisely the case at Mizuho. Every mainframe customer should analyse their R4HA profile and workload contribution patterns as the first step in any licensing optimisation initiative.
Mizuho's batch scheduling had never been analysed from a licensing perspective because batch scheduling was an operations discipline and licensing was a procurement discipline. These two functions operated independently, with no awareness that their decisions were directly linked through IBM's MLC pricing model. Connecting operations and procurement — creating shared visibility into how workload timing affects software costs — is a prerequisite for effective mainframe optimisation.
The way high-cost IBM products are allocated across LPARs has a direct and material impact on sub-capacity licensing costs. LPAR configuration decisions should be made with licensing implications as a primary consideration — not as an afterthought. This requires IBM licensing expertise to be present in infrastructure planning discussions, not only in post-hoc compliance reviews.
In any mainframe environment that has been operating for decades, dormant software products are virtually guaranteed. Products get bundled, installed, and forgotten — but their MLC and support costs continue indefinitely. A systematic product usage review should be conducted at least annually, with a specific focus on identifying products that generate costs but no longer deliver business value.
IBM's commercial team will always attempt to negotiate renewals based on the current (unoptimised) consumption baseline — because this produces the highest pricing. Enterprises should complete their optimisation programme first, then enter renewal negotiations with a demonstrably reduced consumption baseline. The optimised baseline is the enterprise's strongest negotiation lever — it establishes the actual value of IBM's software to the business, rather than the inflated value created by inefficient workload management and configuration.
IBM mainframe licensing costs are the single largest technology expenditure for many global banks — and they are almost universally higher than they need to be. The MLC pricing model creates a direct mathematical relationship between workload scheduling, LPAR configuration, and software costs — but in our experience, fewer than 10% of mainframe customers have ever analysed this relationship systematically. Mizuho's $71 million in savings was achieved entirely through optimisation — rightsizing, realigning, and restructuring — with zero reduction in processing capacity and zero operational risk. The savings were there all along; they just needed to be found.
Situation: A major financial institution sought to reduce IBM mainframe licensing costs ahead of a contract renewal cycle.
Actions: Full MLC analysis, workload optimisation, LPAR restructuring, and renewal negotiation support.
Result: Significant mainframe licensing cost reduction.
Read the full financial institution case study →Situation: Samsung required a comprehensive IBM licensing assessment across its global enterprise to identify optimisation opportunities.
Actions: Full IBM estate assessment, entitlement reconciliation, and optimised licensing strategy.
Result: $23M in verified savings.
Read the full Samsung case study →Situation: A US industrial manufacturer received a $32M IBM audit claim driven by flawed Big Four assumptions and ILMT gaps.
Actions: Audit report deconstruction, ILMT remediation, entitlement recovery, and strategic settlement.
Result: $32M → $1.3M (96% reduction).
Read the full Pennsylvania Manufacturer case study →Our IBM mainframe licensing costs had been treated as a fixed, non-negotiable expense for years. Redress Compliance demonstrated that the costs were not fixed at all — they were driven by workload scheduling and configuration decisions that had never been analysed from a licensing perspective. The $71 million in savings was achieved without reducing any processing capacity and without any impact on our operations. The optimisation framework Redress provided ensures we maintain these savings going forward.
Monthly License Charge (MLC) is IBM's dominant pricing model for z/OS mainframe software. Costs are calculated based on the highest rolling 4-hour average of MSU (Million Service Units) consumption during each month. Because a single 4-hour peak determines the entire month's cost, unmanaged workload scheduling — where batch processing coincides with online transaction peaks — can inflate MLC costs by tens of millions of dollars annually for large environments.
Yes. The optimisations in Mizuho's case changed when workloads execute and how capacity is allocated across LPARs — not whether workloads execute or how much capacity is available. The total processing capability remained unchanged. The savings resulted from aligning workload scheduling and LPAR configurations with IBM's MLC pricing mechanics — an optimisation that is entirely invisible to the applications and end users that depend on the mainframe.
For an environment of Mizuho's scale and complexity (multiple zSeries complexes, global data centres, dozens of IBM products), a comprehensive programme typically takes 12–16 weeks. Smaller environments with fewer mainframe complexes can be optimised in 8–12 weeks. The timeline includes assessment, strategy development, implementation planning, phased execution, and contract negotiation.
MLC peak management savings are sustainable as long as the optimised workload scheduling is maintained. Redress provides an ongoing management framework — including monitoring, guidelines for new workload placement, and change management procedures — that enables the customer's operations team to maintain the optimised profile. LPAR and dormant product savings are permanent structural changes that do not require ongoing management.
Always before. IBM will negotiate based on your current (unoptimised) consumption baseline if you have not yet reduced your MLC profile. Completing optimisation first gives you a demonstrably lower consumption baseline — which is your strongest negotiation lever. In Mizuho's case, the contract restructuring savings ($12M) were only possible because the preceding optimisations had established a reduced usage baseline that justified renegotiating commercial terms.
Redress Compliance's IBM Mainframe Cost Optimisation Programme has delivered hundreds of millions of dollars in verified savings for global financial institutions. The savings are in your workload scheduling, LPAR configurations, and contract terms — we find them.