Beta · Behavioural proof for COBOL-to-Java · SaaS or On-Premise

Don't trust the migration.
Prove it.

ilumeny Verify runs the original COBOL program and the migrated Java service side by side, on the same input and the same database, and compares what really matters: return codes, output files, database state, MQ messages and CICS screen dialogs step by step. Every deliberate difference becomes a documented decision. The result is not a green build, but evidence an auditor can follow.

Talk to us about Verify See how it works
4
program kinds: batch, subprogram, MQ, CICS online
7/7
generated programs of our reference estate proven
0
lines of code leave your premises with the on-premise runner

The question every migration board asks: does it still do the same?

Code reviews and unit tests show that the new code looks right. Verify shows that it behaves like the old one, on real data, with the original as the reference.

⚖️
The original is the specification
The COBOL program runs unchanged in an isolated runner, with a database seeded from the same data as the Java side. No hand-written expectations that can be wrong themselves.
🔍
Compared where it matters
Return code, every output file record, the database after the run, MQ messages put and backed out, and for CICS transactions the HTTP status and data of every dialog step, including the PUT that changes the database.
🧭
Branch coverage of the original
You see which IF and EVALUATE branches of the COBOL program the test data reached, and which not. Gaps are named, so the next test case closes them instead of guessing.
📝
Decisions, not exceptions
An intended difference (a fixed legacy bug, a modern date format) is accepted once, with reason, business reference and approver. The test uses only accepted decisions, and the audit trail keeps every change.
🗂️
Legacy defect catalogue
Known defect types of old code, such as PERFORM ranges left by GO TO, keys never read from EIBAID or items that were never defined, are recognised and answered once in a migration policy, instead of being argued program by program.
📄
Evidence report for audit
One report per project: verdict per program, coverage, decisions with approvers, autopilot class A to D. Built for internal audit and supervisory reviews under MaRisk, BAIT and DORA.

Same input, same database,
two programs, one verdict

01

The original runs as it is

The runner compiles the COBOL program with its copybooks and embedded SQL and runs it against PostgreSQL. CICS and MQ calls go to built-in stubs that record every screen, message and commit.

02

The Java service gets exactly the same

Same input files, same database snapshot, same messages. For online transactions the screen dialog is replayed as REST calls, step by step.

03

Compare and explain

Every difference is shown in context: which record, which column, which dialog step, plus the branches of the original the run reached.

04

Decide or repair

Accept an intended difference, or let the deviation flow back into the code generation as a repair round. Within the ilumeny platform, the next run starts automatically.

05

Evidence, not opinion

VERIFIED, VERIFIED with documented decisions, or open, with the reason. A pipeline quality gate keeps unproven code out of the delivery.

verify — DEPOT101 (CICS, read-write)
→ COBOL: 14 dialog steps, transaction DPT1
→ Java: 12 API calls (11 GET, 1 PUT)
→ SCREEN: status and data of every step · equal
→ DEPOT: 3 / 3 rows · equal
→ AUDIT_LOG: 4 / 4 rows · equal
→ coverage: 45 of 63 branches of the original
 
✓ VERIFIED · documented: 9 policy decisions, 0 open
Evidence report: evidence-report.html
 
$

Inside the platform, or on its own

Verify does not care who wrote the Java. It needs the COBOL original, the Java service and test data.

🔗
Built into ilumeny Transform and Autopilot
Every generated service comes with its test configuration. The differential test runs after each generation, and the verdict feeds the quality gate of the migration pipeline.
🧩
Standalone for your own migration
Already migrating with your own team, a system integrator or another tool? Verify proves that result too. We connect your Java services to the original programs in a fixed-scope pilot.
🏦
On-premise runner
The runner ships as a container image for Docker or Kubernetes in your data centre. Source code, test data and results stay inside your network. EU-hosted SaaS is available as well.

Available standalone, now in Beta

Licensed per project or as part of ilumeny Transform. Start with a pilot on one of your programs.

ilumeny Verify
Standalone module · SaaS or On-Premise
Pricing on request
Get in touch for current Beta conditions, or start with a fixed-scope pilot on your own programs.
  • ✓ Batch, subprograms, MQ and CICS online programs
  • ✓ Return code, output files, database state, messages, screens
  • ✓ Branch coverage of the original
  • ✓ Decision registry with approvals and audit trail
  • ✓ Legacy defect catalogue and migration policy
  • ✓ Audit-ready evidence report
  • ✓ On-premise runner (Docker / Kubernetes) or EU-hosted
Talk to us

Show your auditors
that it behaves the same.

ilumeny Verify is currently in Beta. Tell us about your migration, or let us prove one of your programs in a fixed-scope pilot.

Talk to us about Verify
Questions? Write to hello@ilumeny.com