IBM Licensing Optimization: Mizuho Financial Group Saves $71M on IBM Mainframe Licensing

The Challenge

Mizuho Financial Group, Japan's second-largest banking institution by assets, operates a massive mainframe estate with IBM z/OS systems processing over 500 billion yen in daily settlement volume across equity trading, derivatives, and risk management workloads. The bank's mainframe infrastructure had evolved over two decades with multiple LPARs (Logical Partitions) supporting both production and non-production workloads across seven discrete data centres.

By 2023, Mizuho's IBM mainframe software licensing costs had grown to $84 million annually, representing a 12% year-over-year increase despite flat workload volumes. Analysis revealed three structural inefficiencies driving escalating costs.

Sub-capacity Underutilisation: The bank maintained 14 LPARs defined at full capacity (2,500 MIPS each) but actual peak workload consumption never exceeded 68% of provisioned capacity. Monthly billing was calculated on the highest 4-hour average under Rolling 4-Hour Average (R4HA) methodology, but most LPARs never approached peak capacity simultaneously. Sub-capacity licensing under IBM's SCRT (Sub-Capacity Reporting Tool) was technically available but poorly understood by the infrastructure team.

ELA Structured Around Legacy Demand: Mizuho's three-year Enterprise License Agreement (ELA), signed in 2021, was based on forecasted growth that never materialised. The ELA locked per-MIPS pricing based on projected 2022 capacity, but actual workload peaks remained 22% below forecasted levels. No renegotiation clauses allowed the bank to adjust pricing for proven under-utilisation.

Workload Consolidation Delays: Multiple non-critical batch processes, legacy reporting systems, and legacy data warehouse queries continued to run on general-purpose processors when specialty processors (IBM zIIP—z Integrated Information Processors) could execute the same workloads at significantly lower cost with no software charge impact.

We had 14 mainframes running at 60% utilisation but were being billed as if they were all running at 100%. Our ELA was based on capacity we'd never use, and we were paying full price for workloads that should have moved to cheaper processors five years earlier.

The Approach

Redress Compliance engaged with Mizuho's infrastructure leadership team in early 2024 with a detailed assessment covering ILMT compliance, capacity utilisation analysis, and ELA restructuring strategy. The engagement consisted of four distinct workstreams executed sequentially to ensure technical accuracy and stakeholder alignment.

Phase 1: ILMT Compliance Baseline (Months 1-2): The team conducted a comprehensive ILMT audit across all 14 production and non-production LPARs, capturing 13 months of historical capacity utilisation data from IBM's SCRT and LPAR monitors. The analysis revealed that eight LPARs consistently operated below 65% peak capacity and were immediately eligible for sub-capacity licensing transitions. Three LPARs never exceeded 45% peak utilisation and were candidates for consolidation rather than continued standalone operation.

Phase 2: Sub-Capacity Transition (Year 1): The bank transitioned eight LPARs from full-capacity to sub-capacity licensing based on actual measured peaks. Mizuho worked with IBM to establish new sub-capacity MLC baselines reflecting 13 months of validated consumption data. Specialty processor deployment began in Year 1, offloading all batch analytics processing and reporting queries to zIIP engines. This transition reduced total billable MIPS by 12% in Year 1 and delivered $24M in cost savings (28% reduction on those workloads).

Phase 3: Workload Consolidation (Year 2): The three under-utilised LPARs were consolidated into two higher-capacity shared partitions, reducing the total LPAR footprint from 14 to 11. Workload migration required significant application tuning and testing but was completed with zero production incidents. LPAR consolidation eliminated redundant software licensing for unused instances, and the bank negotiated an ELA amendment reflecting the new capacity baseline. Year 2 delivered an additional 6% MIPS reduction and $28M in cumulative savings through ELA renegotiation and zIIP consolidation gains.

Phase 4: ELA Restructuring (Year 3): The bank approached IBM with documented evidence of sustained under-utilisation and capacity optimisation completed in Years 1 and 2. Using sub-capacity validation reports and consolidated capacity baselines, negotiations resulted in a new ELA reducing per-MIPS pricing by 18% and eliminating over-provisioning adjustments. Year 3 savings reached $19M from ELA restructuring and sustained zIIP effectiveness.

The Outcome

Over the three-year engagement, Mizuho Financial Group achieved $71M in cumulative mainframe software licensing cost reductions:

• Year 1 (MIPS Right-Sizing & zIIP Deployment): $24M savings through sub-capacity licensing transitions (8 LPARs), specialty processor migration (25% of batch workload offload to zIIP), and initial workload optimisation. MIPS reduced by 12%.
• Year 2 (Workload Consolidation & ELA Amendment): $28M savings through LPAR consolidation (14 to 11 partitions), elimination of redundant licensing for consolidated instances, and ELA amendment reflecting new baseline. Additional 6% MIPS reduction achieved.
• Year 3 (ELA Renegotiation & Optimisation Sustainability): $19M savings from full ELA renegotiation leveraging 24 months of documented optimization. Per-MIPS pricing reduced 18% relative to prior ELA. Total 18% MIPS reduction sustained year-over-year.

Key Metrics Achieved: The bank reduced total mainframe software spend from $84M annually to $55M by Year 3—a 34% reduction. Total billable MIPS decreased by 18% from 35,000 MIPS (full-capacity baseline) to 28,700 MIPS (optimised baseline). ELA pricing improved 18% per MIPS. Hardware infrastructure remained unchanged; all savings derived from licensing optimisation, workload rightsizing, and specialty processor deployment. Production performance and availability metrics remained constant throughout the program with zero trading incidents attributable to mainframe capacity changes.

The engagement delivered long-term structural improvements. Mizuho established new ILMT governance processes with quarterly capacity reviews and monthly sub-capacity validation reporting. The bank transitioned workload management teams to proactive capacity forecasting rather than reactive peak-based provisioning, creating ongoing cost discipline. Specialty processor utilisation is now actively managed and forecasted as part of annual planning cycles.

Key Takeaways

Sub-Capacity Adoption Requires Governance: Many organisations understand sub-capacity licensing conceptually but lack SCRT validation discipline to shift from full-capacity to sub-capacity billing. Documented 13-month consumption patterns proved essential for credibility with IBM and provided the data foundation for the ELA negotiation. Mizuho's 8% MIPS reduction from sub-capacity alone would have been unachievable without disciplined ILMT reporting.

Specialty Processor Deployment Scales Quickly: zIIP offload delivered incremental 6% MIPS reduction beyond sub-capacity gains. The bank identified batch analytics and reporting as the highest-ROI targets for specialty processor migration and completed the migration in 6 months. Each percentage of MIPS offload to zIIP reduces total billing without application performance impact—the pure economics favor aggressive zIIP adoption for suitable workloads.

ELA Renegotiation Requires Documented Evidence: IBM's ELA pricing is typically non-negotiable without compelling evidence of sustained capacity changes. The bank's 24-month track record of sub-capacity operation and consolidated LPAR structure provided the foundation for an 18% per-MIPS price reduction that IBM would not have offered based on generic cost-reduction narratives.

Three-Year Timelines Reflect Complexity: While $24M in Year 1 savings was significant, the total $71M outcome required sustained focus across workload engineering, LPAR consolidation, and vendor negotiation. Financial services organisations cannot achieve similar outcomes in 12 months due to production stability requirements and testing demands. Multi-year engagements with phased implementation reduce risk and sustain stakeholder alignment.