U.S. Air Force ILS-S: Unisys COBOL to Java on AWS GovCloud
How TSRI, CGI and NTT DATA modernized “The Beast”: the 54-year-old Standard Base Supply System behind daily supply support for Air Force missions.
Executive summary
The quintessential mission-critical system
TSRI was contracted by CGI Federal under the NETCENTS-2 contract vehicle to help modernize the Integrated Logistics Systems – Supply (ILS-S). More than 20,000 users at 280 U.S. Air Force, Air Force Reserve and Air National Guard bases and forward operating locations use ILS-S for daily supply and equipment support for wartime flying missions.
Part of ILS-S was a modern Java/Oracle® application, but the Standard Base Supply System (SBSS), its core business logic and authoritative data, ran on Unisys® mainframes. The modernization targeted AWS GovCloud (US) to meet the Air Force’s stringent security and compliance requirements.
The goal: equal or better performance and functionality on AWS GovCloud (US), with little to no code freeze and full documentation after 50+ years of development.
01 · History
A mission-critical system on a Unisys mainframe
SBSS provides the core business logic and authoritative data for base-level management of supplies and equipment for warfighting missions. The project’s core requirement: the modernized SBSS had to perform equivalently to, or better than, the legacy system. The Air Force selected AWS GovCloud (US) because it aligned with the future technical direction of ILS-S, including cybersecurity, DevOps and AWS’s native continuous integration and delivery tools.
02 · Selection
Why Team CGI chose TSRI
Team CGI reviewed a total rewrite, a re-architecture, a COBOL-to-Micro Focus COBOL migration, and automated code conversion. A rewrite or re-architecture failed the program’s time constraints, had historically low success rates and cost too much; the Micro Focus migration was a stopgap that didn’t reach the Red Hat Enterprise Linux/Java/Oracle future state. Automated code conversion had the highest probability of success in a cloud-native environment at the lowest cost.
Team CGI reviewed 12 code conversion vendors, then ran a competition among four, and selected TSRI because:
TSRI had converted 300,000 lines of COBOL to C++ and Java/JEE for the Air Force REMIS program, earning the prime contractor’s “Small Business of the Year.”
Its automated tooling could convert the legacy SBSS code with the highest probability of meeting the 25-month program schedule.
Its solution had the lowest total price of the tools assessed.
Its blueprints gave direct 1-to-1 COBOL-to-Java traceability, confirming every capability migrated and identifying Unisys operating system gaps.
03 · Technology
A model-driven engine with a long Air Force track record
At the core of TSRI’s technology is Janus Studio®, an automated, model-driven transformation engine. Rather than transliterating source to target, it first builds a comprehensive model of the legacy system in an intermediate language, enabling fully automated refactoring between source and target languages and code-level documentation.
The Application Blueprint® and Transformation Blueprint® let developers study the as-is COBOL and the transformed Java side by side, with control flow, data flow, cause-effect, complexity analysis and state transition tables.
04 · SBSS modernization
Modernizing “The Beast”
SBSS comprised 1,260,679 lines of Unisys COBOL and 10,078 lines of C. It had existed for more than 54 years, and earlier attempts to move it off Unisys had failed so completely that the book Modernizing Legacy Systems referred to it as “The Beast.” The target was a mid-tier Java environment with Unix shell scripting, Red Hat Enterprise Linux®, Apache web servers, JBoss and Oracle RDBMS on AWS GovCloud (US).
05 · Refactoring
From 7.9 million lines of Java to 5.2 million
The Unisys network database code was first transformed into 1.7 million lines of COBOL with SQL, then to Java with the rest of the application. TSRI’s automated refactoring (identical record consolidation, DAO method consolidation, unused definition removal) and an improved method synthesis algorithm brought 7.9 million lines down to 5.2 million.
- 01 · SPIRAL
Regenerate
The full code base, every iteration
- 02 · DELIVER
Compiling Java
Integration-ready, externals stubbed
- 03 · INTEGRATE
ARRAY team
Schedulers and utilities reintegrated
- 04 · FEEDBACK
Defects via Bugzilla
Rules adapted and code regenerated
06 · Testing
Proving functional equivalence with telemetry
TSRI injected telemetry into the legacy COBOL system in production, capturing inputs and outputs to compare with the modernized system. Discrepancies fed straight back into TSRI’s toolset, enabling faster debugging and proof of equivalence. Of 290 issues found, many were latent bugs in the original code; only 62 came from transformation, one per 21,000 lines, against industry norms of 15–50 defects per 1,000 lines written manually.
07 · Lessons learned
Extending Janus Studio® for SBSS
ILS-S used COBOL features Janus Studio® had not yet defined, and control flow with many GOTOs jumping out of perform ranges. TSRI characterized the behaviors on the original mainframe, updated its frameworks, and extended its GOTO elimination strategies.
08 · Results
Deployed to AWS GovCloud
The transformed application was deployed to AWS GovCloud (US), and the project was considered highly successful. In parallel, the team delivered three major FIAR releases, migrated into big data and began a DoD-leading mobile implementation. Janus Studio®’s tuned rules carry forward to future Air Force COBOL projects.
Unisys COBOL and C
1,260,679 lines of COBOL and 10,078 lines of C transformed to Java.
From transformation
One per 21,000 lines, versus 15–50 per 1,000 lines written manually.
Of refactored Java
Down from 7.9 million after the first pass.
AWS GovCloud (US)
Meeting the Air Force’s security and compliance requirements.
The automated Janus Studio® toolset learns with each project, and the transformation rules it creates are reused in subsequent projects.