Key Takeaways: Migration and Modernization

  1. The modernization spectrum has five positions: Maintain, Wrap, Extend, Re-Architect, and Replace. Each has different cost, risk, and reward profiles. Most organizations should start on the left (lower risk) and move right only when justified.

  2. API wrapping is the most successful modernization approach. It exposes existing COBOL functionality through REST APIs without changing the underlying code. z/OS Connect and custom wrappers handle the JSON-to-COMMAREA transformation.

  3. "Legacy" does not mean "obsolete." A system is legacy when it no longer meets business needs, not simply because it is old or written in COBOL. Many COBOL systems are mature, reliable, and cost-effective.

  4. Rewriting is the riskiest strategy. Industry data shows approximately 50% of large-scale rewrite projects fail. The Strangler Fig pattern mitigates this risk through incremental replacement.

  5. COBOL can run in the cloud. GnuCOBOL compiles to native Linux executables. Micro Focus provides managed runtimes. Containerized COBOL programs can run in Docker and Kubernetes.

  6. COBOL microservices are viable. A COBOL program that performs a single business function, compiled with GnuCOBOL and wrapped in a REST interface, fits the microservice pattern perfectly.

  7. Database migration requires program changes. Converting from VSAM or IMS to DB2 means replacing file I/O statements with embedded SQL — a significant code change that must be thoroughly tested.

  8. AI assists but does not decide. AI tools can summarize code, extract business rules, and suggest refactoring, but they cannot understand business context or make strategic decisions.

  9. The risk of doing nothing is real. Developer talent gaps, rising costs, and capability limitations create increasing business risk. Doing nothing is itself a choice with consequences.

  10. Different subsystems deserve different strategies. The best modernization plans evaluate each component independently and place it at the appropriate point on the spectrum — not one-size-fits-all.