Book a demo

Canada Revenue Agency Helpdesk: IBM COBOL to Java on AWS

How TSRI and AWS modernized the CRA Helpdesk application, from mainframe COBOL, CICS and green screens to cloud-native Java, Angular and PostgreSQL, transforming all source code in under three days.

On time and within budget
Code warranty
No license fees
CLIENTCanada Revenue Agency
SOFTWAREHelpdesk application
LANGUAGE PAIRINGIBM COBOL → Java
TIME TO COMPLETION7.5 months

Executive summary

Mainframe to the cloud

CRA and AWS engaged TSRI to modernize CRA’s COBOL-based system to Java on AWS. TSRI tuned its Janus Studio® transformation engine to parse, document, transform and refactor the IBM z/OS COBOL application.

TSRI modernized and refactored the CRA Helpdesk application from a monolithic IBM mainframe COBOL and CICS application to a modern, multi-tier Java architecture. All BMS green screens were automatically transformed to Angular web screens, a REST layer was introduced, and DB2 and VSAM sequential files were converted to PostgreSQL on Amazon Aurora.

TSRI then completed integration and testing and deployed the containerized application on AWS.

01 · Challenge

Two problems that felt insurmountable

The Canada Revenue Agency (CRA), Canada’s tax collection and enforcement agency, faced two key problems: operational costs to maintain its mainframe systems had spiraled out of control, and finding people to keep programming the agency’s legacy applications had become harder with each passing year.

CRA had been planning mainframe modernization for its IBM portfolio for some time, but a high-risk COTS replacement or manual rewrite could take many years and carry a high risk of failure. To reduce operating costs, find and retain people with the right skills, and be more agile, it needed to modernize its COBOL, JCL, user interface and DB2 database.

02 · Selection

Why AWS and CRA chose TSRI

Before launching a full project, CRA and AWS needed proof that automated modernization would accelerate migration off the mainframe and lower risk. CRA investigated multiple migration patterns (rewrite, rehost, replatform, lift-and-shift and COTS replacement) and reviewed multiple modernization vendors.

CRA chose TSRI in part because TSRI was a 2021 launch partner for the AWS Mainframe Modernization competency, for its unique refactoring capability, and for its ability to modernize COBOL, JCL, DB2 and VSAM, and BMS green screens to Java, Python and an Angular web front end.

This team effort used AWS Migration Acceleration Program (MAP) funding, TSRI’s Mainframe Modernization Competency, and AWS ProServe and CRA resources.

MAP: Assess, Mobilize, Migrate & Modernize

Funding and resources from AWS

Janus Studio® tuned for CRA

Reusable for future CRA projects

03 · Approach

Compiling Java in days, not years

TSRI tuned Janus Studio® specifically for CRA before any code was modernized. Within the first 3 days, TSRI had modeled and transformed all code to compiling Java; on day 10 it delivered an Application Blueprint® documenting the legacy Helpdesk application in fully linked hypertext.

How the project ran

  1. 01 · DAY 3

    Compiling Java

    All COBOL modeled and transformed

  2. 02 · DAY 10

    Application Blueprint®

    Fully linked legacy documentation

  3. 03 · DATA

    Schema & ETL

    PostgreSQL schema, test data loaded

  4. 04 · DAYS

    Frameworks on AWS

    Deployed to the AWS landing zone

  5. 05 · REFACTOR

    SonarQube-guided

    Automated and customer-driven

What changed, layer by layer

Layer
Before
After
Application
Monolithic IBM COBOL + CICS
Multi-tier, cloud-native Java
User interface
BMS green screens
Angular, CSS and HTML5 web front end
Interfaces
CICS
REST layer and a functionally equivalent Java framework
Data
DB2 and VSAM sequential files
PostgreSQL on Amazon Aurora
Hosting
IBM mainframe
Containers on AWS, Well-Architected model

Blueprints that cover the whole system

The Application Blueprint® covered not just the COBOL code, but the CICS interfaces, externals, hundreds of screens and connected utilities. A Transformation Blueprint® showed the modernized Java codebase and how DB2 and VSAM data would move to PostgreSQL through TSRI’s dual database layer.

application blueprint · includes
Package structureArchitecture design metricsExternal interfacesCRUD indicesControl flow chartsEmbedded SQL & CICSStructure chartsDead codeCall fan-in & fan-outSimilarity analysisData element tablesClass diagramsState machine modelsState transition tablesModule diagramsCause-effect graphsCyclomatic complexityDependency analysis

04 · Architecture

CRA Helpdesk on AWS

The modernized solution was containerized and deployed on the provided AWS landing zone with native AWS services and the TSRI framework. All code, including the framework, was delivered as source code, with a code warranty and no license fees.

CRA HELPDESK · AWSSIMPLIFIED
ACCESS
End usersElastic Load Balancing (ALB)Amazon Cognito & IAM*AWS Directory Service*
CONTAINERS · ECS ON FARGATE
Presentation servicesApplication servicesTSRI Client Framework LibraryTSRI DAOCICS framework library
SESSION
Redis cache
DATA
PostgreSQL on Amazon AuroraAWS DataSync*
OPERATIONS
Amazon CloudWatch
Simplified from the original CRA Helpdesk architecture diagram; connections are not shown. * CRA IT dependent.
01

Wrapping CICS for the cloud

A CICS gateway lets many transactions run at once, essential for an agency that handles thousands of transactions a day and millions around late April and quarterly corporate tax deadlines. TSRI built a REST interface around the CICS interfaces, separating the presentation layer from the container layer so the application can scale through peak periods on AWS’s Well-Architected model.

02

Working through pandemic constraints

COVID-19 delayed some materials TSRI needed, and subject-matter experts weren’t always available to explain screen-by-screen flow, which slowed validation across hundreds of BMS screens. Limited access to highly secure areas created further hurdles, which TSRI worked around.

05 · Refactoring

From functional equivalence to high-quality code

After reaching functional equivalence, TSRI used fully automated, semi-automated and customer-driven refactoring to meet CRA code specifications and a target SonarQube quality score: removing dead code, consolidating similar records and code, adapting the package structure, and remediating SonarQube issues. CRA supplied updated legacy baselines instead of freezing code.

99.982%automated

COBOL to Java

All mainframe source transformed in less than 3 days.

<8bugs

From thousands of issues

All latent defects in replacement services and CICS implementations, and all remedied.

0defects

Transformation-related

Across the full transformation.

~80%projected

Lower annual TCO

From the AWS business case, once the full mainframe architecture is migrated.

06 · Results

On time, on budget, and ready to scale

With AWS MAP funding, CRA, TSRI and AWS completed the project on time and within budget. The code and data transformation took about three weeks; the rest of the schedule went to externals and utilities, testing, integration, deployment and customer-driven refactoring. TSRI and AWS are now working with CRA to apply the same tuned rules to the rest of its application portfolio.

Get modern fast.
Stay modern forever.

See Juno in action!