Chapter 34 Exercises: COBOL-to-Cloud Patterns

Part A: Conceptual Questions

A1. Explain the difference between allocated MIPS cost and marginal MIPS cost. Why does this distinction matter for COBOL-to-cloud TCO analysis? Give a specific example where removing 500 MIPS of batch workload from a 6,000-MIPS installation produces zero mainframe cost savings.

A2. The chapter identifies I/O architecture as "the single most underestimated factor in COBOL-to-cloud migrations." Explain why the z/OS I/O subsystem (FICON channels, data spaces, LSR buffering) produces fundamentally different I/O behavior than cloud block storage (io2 Block Express, Ultra Disk). Why can't you close this gap by buying more IOPS?

A3. Compare and contrast the three rehosting platforms covered in Section 34.2: Micro Focus Enterprise Server, Heirloom Computing, and NTT DATA UniKix. For each, identify one scenario where it is the best choice and one scenario where it is a poor choice. Justify your reasoning based on the technical characteristics, not cost alone.

A4. Sandra Chen's FBA eligibility verification pilot showed a p99 latency increase from 8.4ms to 340ms when moved to AWS GovCloud. Decompose this latency increase into its contributing factors: CICS emulation overhead, network latency between application and database, database buffer hit ratio change, data conversion overhead. Which factor contributed the most, and why?

A5. The chapter states that "Parallel Sysplex data sharing has no cloud equivalent" and calls this a "hard no." Explain specifically what the Coupling Facility provides (lock structure, group buffer pool, list structure) and why distributed software locks over a network cannot replicate this. What is the fundamental physical constraint?

A6. Describe the EBCDIC-to-ASCII conversion problem in detail. Give three specific examples of COBOL constructs or data patterns that cause subtle errors during conversion: (a) a hardcoded hex literal, (b) a collating sequence dependency, and (c) a packed decimal sign nibble issue. For each, explain how the error manifests and how it can be detected.

A7. Explain why moving dev/test environments to the cloud is "almost always the right answer." Address the objection: "But if our dev/test environment doesn't perfectly match production, we'll miss bugs." How does the chapter respond to this concern?

A8. Rob Calloway spent 340 hours in Year 1 managing the cloud reporting environment. This is described as "opportunity cost" because it delayed his mainframe batch window optimization project. Explain why opportunity cost should be included in TCO analysis. How would you quantify it? Should it be a line item in the spreadsheet, or should it be handled differently?

Part B: Applied Analysis

B1. Cloud Candidacy Assessment

A regional insurance company runs the following COBOL workloads on an IBM z15 with 4,500 MIPS. Evaluate each workload for cloud candidacy using the decision tree from Section 34.7. For each, provide a verdict (Cloud, Mainframe, Hybrid, or Evaluate Further) with specific justification.

For each workload: 1. Apply the three questions from Section 34.7 2. Identify the primary workload characteristic 3. Estimate marginal mainframe savings potential (High/Medium/Low/Zero) 4. State your verdict with reasoning 5. Identify the biggest risk if your recommendation is wrong

B2. Rehosting Platform Selection

Using the compatibility matrix from Section 34.2, recommend a rehosting platform for each of the following migration scenarios. Justify your choice.

Scenario A: A state government agency wants to move its batch-only tax processing system (220K LOC COBOL, no CICS, DB2 for data storage, JCL-scheduled jobs with GDG dependencies) to AWS GovCloud. No real-time requirements. Cost is the primary driver.

Scenario B: A mid-size bank wants to move its internal branch operations system (85K LOC COBOL, CICS with BMS maps, DB2, ~400 TPS during business hours) to Azure. The bank wants to use the same operational tools (Azure Monitor, Azure DevOps) for mainframe and cloud workloads.

Scenario C: A retail company wants to eliminate its mainframe entirely within 3 years. It has 500K LOC COBOL across CICS, batch, and IMS. The CTO wants the resulting code to be maintainable by Java developers.

For each scenario: 1. Recommended platform and why 2. Components that will work well 3. Components that will require significant remediation 4. Key risk and mitigation strategy 5. Estimated additional effort (beyond the vendor's initial estimate) as a percentage

B3. TCO Calculation Exercise

Using the following data, build an honest 5-year TCO comparison for migrating a COBOL batch analytics workload to cloud.

Current mainframe costs: - Workload consumes 600 MIPS - Runs during off-peak hours (02:00-05:00), does NOT contribute to R4HA peak - Fully allocated cost: 600 × $5,500/MIPS = $3,300,000/year - Marginal mainframe cost reduction if workload is removed: $180,000/year (sub-capacity relief on one software product) - 1.5 FTE maintaining the system: $255,000/year (these people are NOT freed up — they'll maintain other systems)

Projected cloud costs (from vendor proposal): - EC2 compute (reserved, 3-year): $38,000/year - Storage (io2 + S3): $24,000/year - Micro Focus license (4-core): $84,000/year - Data transfer: $6,000/year - Vendor total: $152,000/year - Vendor projected savings: $3,300,000 - $152,000 = $3,148,000/year (95% savings!)

Additional data you've discovered: - Migration project estimated at $380,000 (vendor says $180,000) - Batch window expansion from 3 hours to 7 hours requires larger instances: add $22,000/year - Cloud operations (0.3 FTE new skillset): $51,000/year - Direct Connect for data replication: $36,000/year - Micro Focus maintenance: $16,800/year - Security/compliance (initial audit + annual): $45,000 Year 1, $18,000/year ongoing - Monitoring/alerting tools: $9,000/year - Mainframe cost may actually increase by $42,000/year (pricing model impact — same pattern as CNB)

Calculate: 1. The vendor's claimed 5-year TCO savings 2. Your honest 5-year TCO savings 3. The ratio between them 4. The actual break-even point (in months) 5. Whether this migration passes "Kwame's Rule" (>30% honest savings after all costs)

Part C: Design Exercises

C1. Hybrid Architecture Design

Design a hybrid architecture for a mid-size credit union with the following systems:

For your hybrid architecture: 1. Draw the architecture diagram showing what runs where (mainframe vs. cloud) 2. Specify the data synchronization pattern for each cross-platform data flow (batch extract, CDC, or API) 3. Identify the integration components needed (z/OS Connect, MQ bridge, CDC tool, etc.) 4. Estimate the cloud infrastructure requirements (instance types, storage, network) 5. Identify three failure scenarios and how the hybrid architecture handles each

C2. Data Synchronization Design

For CNB's hybrid architecture (Section 34.5), design the CDC implementation in detail:

1. Which DB2 tables should be replicated via CDC? (Consider: account master, transaction log, customer information, account balance, hold/freeze records)
2. What is the acceptable replication lag for each table? Why?
3. What happens when the CDC agent falls behind during the nightly batch window? Design the recovery procedure.
4. How do you handle DB2 schema changes (ALTER TABLE) without breaking CDC? Design the change management process.
5. How do you validate that the cloud replica is consistent with the mainframe source? Design the reconciliation job.

C3. Cloud Migration Runbook

Write a production cutover runbook for migrating a batch reporting workload from mainframe to cloud. The runbook should cover:

1. Pre-cutover validation steps (what must be true before you start)
2. Cutover sequence with rollback decision points
3. Parallel run procedures (both mainframe and cloud produce output; compare)
4. Validation criteria (how do you know the cloud output is correct?)
5. Rollback procedure (if validation fails, how do you revert to mainframe-only?)
6. Post-cutover monitoring (first 24 hours, first week, first month)
7. Mainframe decommissioning criteria (when is it safe to turn off the mainframe batch jobs?)

Part D: Critical Thinking

D1. A vendor presents a proof-of-concept showing their COBOL rehosting platform running a CICS transaction at 500 TPS on AWS. The POC used a 1,000-record database. Your production database has 50 million records. Identify five specific reasons why the POC performance will not hold in production. For each reason, explain the technical mechanism.

D2. An organization's CTO says: "In five years, cloud performance will match mainframe performance for all workloads. We should plan our migration now." Evaluate this claim. Consider: Moore's Law trends for x86, z/Architecture investment trajectory, the Coupling Facility hardware, I/O subsystem architecture, and the software ecosystem (CICS, DB2, WLM). Is the CTO's timeline reasonable? What would need to change?

D3. The chapter presents "Kwame's Rule" — don't migrate if honest TCO savings are less than 30%. A junior architect argues that this rule is too conservative: "Even 15% savings is worth it if we also get agility benefits like faster deployment and modern tooling." Construct both sides of this argument. Under what conditions is the junior architect right? Under what conditions is Kwame right?

D4. Sandra Chen ran a six-week OLTP pilot before committing to keep eligibility verification on the mainframe. The pilot cost approximately $120,000 (infrastructure + staff time). A colleague argues this was wasted money: "We could have analyzed the requirements and decided without a pilot." Argue why the pilot was worth $120,000 despite confirming what Sandra already suspected. What did the pilot produce that analysis alone could not?

D5. Consider the ethical dimension: a consulting firm presents a COBOL-to-cloud TCO analysis using allocated MIPS costs (the "$5,000/MIPS" number) instead of marginal costs. Is this an honest mistake or a misleading presentation? What professional obligation does a consultant have to explain the difference? Draft a two-paragraph response you would give to a CTO who asks "Is this TCO analysis trustworthy?"

Part E: Research and Investigation

E1. Research the current pricing for Micro Focus Enterprise Server (now OpenText) on AWS. Find the per-core license cost for: (a) base COBOL runtime, (b) CICS emulation (Enterprise Server CICS), (c) JCL interpreter, and (d) full Enterprise Server suite. Compare with the numbers cited in the chapter. How have they changed?

E2. Research IBM's current Tailored Fit Pricing models for z/OS. Explain: (a) the Enterprise Consumption Solution, (b) the Enterprise Capacity Solution, and (c) how each affects the marginal MIPS cost calculation for a workload being considered for cloud migration. Which pricing model makes cloud migration TCO look better? Which makes it look worse?

E3. Research one publicly documented COBOL-to-cloud migration (successful or failed). Sources: Gartner case studies, AWS/Azure/GCP customer stories, conference presentations. Summarize: (a) the organization and workload, (b) the platform chosen, (c) the outcome, (d) the key lessons, and (e) how the outcome compares with the predictions in this chapter.

E4. Research GraalVM's support for COBOL. Can COBOL be compiled to GraalVM native image? What are the implications for cloud deployment (startup time, memory footprint, peak performance)? How does this compare with Micro Focus Enterprise Server's approach?

E5. Research the current state of CICS emulation on cloud platforms. Has the compatibility gap described in Section 34.2 narrowed since this chapter was written? Specifically investigate: (a) channels and containers support, (b) event processing support, (c) MRO emulation, and (d) CICSPlex SM equivalent. Document your findings with version numbers and dates.