Case Study 1: Global Bank Migration — Continental United Bancorp Evaluates Its Database Future

Background

Continental United Bancorp (CUB) is a top-20 American bank holding company headquartered in Charlotte, North Carolina. With $180 billion in assets, 8.5 million customers, and operations in 34 states, CUB is a significant player in retail banking, commercial lending, and wealth management.

CUB's technology infrastructure is built on a foundation of IBM Z mainframes running z/OS. The crown jewel of this infrastructure — the core banking platform known internally as "Granite" — runs on DB2 for z/OS. Granite has been in production since 1991 and has been upgraded through every major DB2 version, most recently to DB2 12 for z/OS at function level 508.

Granite by the Numbers

Granite is, by any measure, a well-run system. But CUB's Chief Technology Officer, Marcus Webb, is under pressure.

The Pressure

Three forces are converging on Webb's office:

1. Cost Pressure

CUB's mainframe software costs — including DB2, CICS, z/OS, and related products — total approximately $45 million per year. The board's technology committee has been reading articles about banks "moving to the cloud" and saving 40% on infrastructure costs. They want a plan to reduce technology spending by $15 million annually within three years.

2. Talent Pressure

Of CUB's 18 DB2 for z/OS DBAs, 11 are over the age of 55. Three have announced retirement within the next 18 months. Recruiting replacements has been difficult — the last two DB2 z/OS DBA positions took 9 and 14 months to fill, respectively. Webb's HR partner has told him bluntly: "We cannot sustain this staffing model."

3. Innovation Pressure

CUB's digital banking team wants to launch new products faster. Today, any change to Granite — a new table, a modified stored procedure, even a column addition — requires a change-management process that averages 6 weeks from request to production deployment. The digital team, accustomed to deploying microservices multiple times per day, finds this intolerable.

The Proposal

Webb commissions a study from a major consulting firm, Accenture-Deloitte-McKinsey (a fictional amalgamation, but the analysis is realistic). After four months and $2.8 million in consulting fees, they deliver a 200-page report with three options:

Option A: Full Migration to PostgreSQL on AWS

• Migrate all 14,200 tables and 6,800 stored procedures to Amazon Aurora (PostgreSQL-compatible)
• Rewrite 2,200 COBOL programs in Java
• Estimated timeline: 5-7 years
• Estimated cost: $380-520 million
• Estimated annual savings after completion: $28 million (mainframe elimination)
• Risk rating: Very High

Option B: Hybrid Modernization

• Keep DB2 for z/OS as the system of record for core banking
• Build new applications on Db2 for LUW and PostgreSQL on distributed/cloud infrastructure
• Implement real-time data replication from z/OS to distributed platforms
• Gradually shift workloads as applications are rewritten, with no hard deadline for mainframe retirement
• Estimated timeline: 3-5 years for initial phase; mainframe coexistence indefinitely
• Estimated cost: $85-120 million
• Estimated annual savings: $8-12 million (reduced mainframe growth, not elimination)
• Risk rating: Moderate

Option C: Optimize in Place

• Stay on DB2 for z/OS, invest in optimization to reduce CPU consumption (and therefore licensing costs)
• Implement zIIP specialty processors to offload Java and XML workloads from general-purpose processors
• Adopt DB2 12's continuous delivery features aggressively
• Invest in training and recruiting to address the talent gap
• Estimated timeline: 12-18 months
• Estimated cost: $15-25 million
• Estimated annual savings: $6-10 million (reduced CPU costs, not mainframe elimination)
• Risk rating: Low

The Debate

Webb presents the three options to his leadership team. The debate is heated.

The CFO favors Option A. "The math is clear," she argues. "We spend $380 million over five years, but we save $28 million every year after that. The payback period is 13.5 years from start, or 8.5 years from completion. And we eliminate our mainframe dependency permanently."

The Chief Risk Officer is alarmed by Option A. "You are proposing to rewrite the system that processes 42 million transactions per day while it is running. The risk of a significant outage during migration is not theoretical — it is near-certain. One major outage could cost us $50-100 million in direct losses and regulatory penalties, not counting reputational damage. We are a bank, not a startup. We do not get to 'fail fast.'"

The Head of Digital Banking supports Option B. "I don't need the mainframe to go away. I just need it to stop being a bottleneck. If we can get real-time data replication to a distributed platform, my team can build new products without waiting 6 weeks for a change to Granite. Keep the mainframe for what it's good at; give us modern tools for everything else."

The VP of Infrastructure notes a factor the consultants underweighted: "Options A and B both require skills we don't have. We need cloud architects, PostgreSQL DBAs, Kafka engineers, and DevOps specialists. The market for those skills is just as competitive as the mainframe market, and the salaries are comparable. The talent problem doesn't go away — it changes shape."

A senior DB2 DBA, invited to the meeting at Webb's insistence, makes a point about the data: "The consultants estimated $380-520 million for Option A. That range — $140 million — is larger than the entire cost of Options B or C. When your estimate has a 37% uncertainty band, it means you don't actually know what it will cost. I have seen migrations like this. They always cost more and take longer than projected. Always."

The Decision

After three weeks of additional analysis, Webb recommends Option B (Hybrid Modernization) with elements of Option C. His plan:

1. Immediate (Year 1): Optimize DB2 for z/OS CPU consumption using zIIP offload, compression improvements, and SQL tuning. Target: $5 million annual savings. Invest in DB2 DBA recruiting and a training program for junior staff.

2. Short-term (Years 1-2): Deploy Db2 for LUW on Linux as an operational data store. Implement real-time replication from DB2 for z/OS using IBM InfoSphere Data Replication. Enable the digital banking team to build new applications against the LUW platform.

3. Medium-term (Years 2-4): Gradually migrate lower-criticality batch workloads from z/OS to LUW. Evaluate each workload individually — some may move to PostgreSQL or cloud-native databases if appropriate.

4. Ongoing: Continue using DB2 for z/OS for core banking transactions. Accept that the mainframe will be part of CUB's infrastructure for the foreseeable future. Adopt DB2 12 function levels as IBM delivers them.

The board approves the plan with the understanding that Option A remains on the table for future evaluation as cloud and open-source database technologies mature.

Discussion Questions

Question 1: The Cost Analysis (Intermediate)

The CFO's payback analysis for Option A calculated a 13.5-year payback from project start (or 8.5 years from completion). What factors might she be underestimating or ignoring in this calculation? Consider at least five hidden costs.

Question 2: Risk Quantification (Advanced)

The Chief Risk Officer asserts that the risk of a significant outage during migration is "near-certain." Is this an overstatement? What risk-mitigation strategies could be employed during a large-scale database migration to reduce (but not eliminate) this risk? How would you quantify the risk in dollar terms?

Question 3: The Talent Dilemma (Intermediate)

The VP of Infrastructure argues that the talent problem "changes shape" rather than going away. Evaluate this claim. Is it easier to hire PostgreSQL DBAs than DB2 for z/OS DBAs? If so, does that advantage offset the loss of institutional knowledge when the mainframe DB2 team retires?

Question 4: The Hybrid Architecture (Intermediate)

Option B creates a dual-platform environment similar to Meridian National Bank's architecture. What are the operational challenges of running both DB2 for z/OS and Db2 for LUW simultaneously? Consider data consistency, staff skills, monitoring, and incident management.

Question 5: The Consultant's Blind Spot (Advanced)

The senior DBA noted that the consultants' $140 million uncertainty band for Option A was larger than the entire cost of Option C. What does this say about the reliability of large-scale migration estimates? What estimation methodologies might produce more reliable figures? Should organizations trust multi-hundred-million-dollar investment decisions to estimates with this level of uncertainty?

Question 6: Your Recommendation (Advanced)

If you were in Marcus Webb's position, would you have chosen the same option? Write a one-page memo to the board defending either the same choice or a different one. Your memo must address cost, risk, talent, and timeline.

Return to Chapter 1 | Continue to Case Study 2