Book a demo

Pitney Bowes Postage Payment Application: COBOL to C# .NET Core on AWS

How TSRI moved a decades-old HP NonStop Tandem COBOL application and its Tandem database to C# .NET Core and Microsoft SQL Server on AWS, with no code freeze or business disruption.

No code freeze
No business disruption
Deployed on AWS
CLIENTPitney Bowes
SOFTWAREPostage Payment System
LANGUAGE PAIRINGCOBOL → C# .NET Core
TIME TO COMPLETION6 months

Executive summary

Mainframe to the cloud

For decades, the Pitney Bowes Postage Payment Application had been running COBOL on an HP NonStop Tandem mainframe. To seize the opportunities of the digital cloud age and reduce overall technical debt, Pitney Bowes needed to modernize to C# .NET Core, transform its Tandem Database to a modern Microsoft SQL Server database, and deploy the modernized application on AWS.

TSRI modernized 390,000 total lines of code to a modern architecture in a transformation that involved no code freeze or business disruption, despite multiple code baseline updates adding over 40,000 lines of code.

The business case developed with AWS projects that the application’s annual Total Cost of Ownership will drop by nearly 97%, from multiple millions of dollars a year to less than $30,000.

01 · The challenge

A century of innovation, running on a 1959 language

In 1920, the Pitney Bowes Postage Meter became the first such device approved for use throughout the entire U.S. postal system. It was the first in a century of innovation, from coin-operated postage machines to continuous navigation systems and fax technology, with communication and postal payment services at the core.

Mail payment remains one of Pitney Bowes’ core businesses, now international and digital. Yet in 2021, its postage payment processing system still ran on COBOL, a language developed for mainframes in 1959.

For a company as innovation-driven as Pitney Bowes, the ability to scale not only horizontally in size, but vertically in capabilities, is key.

Total cost of ownership

Millions of dollars a year.

A shrinking, expensive talent pool

Fewer and more expensive people who still know COBOL.

Hardware compatibility

New hardware is less compatible with mainframes.

Scale in size and capability

The need to grow horizontally and vertically.

Modernization was also a practical continuation of Pitney Bowes’ hundred-year culture of ingenuity.

02 · Solution and process

A returning customer and a proven, model-based process

In 2016, working with Accenture, TSRI modernized Pitney Bowes’ Visual Basic 6 Tier 3 application to C# .NET Core, now also deployed on AWS. TSRI parsed and modeled that application into a legacy Abstract Syntax Tree (AST), transformed it into an Intermediate Object Model (IOM), then generated the modern target language. Based on those outcomes, the choice of TSRI for the move to AWS was clear.

How the project ran

  1. 01 · FIRST DAYS

    Application Blueprint®

    Full, hyperlinked documentation of the legacy COBOL

  2. 02 · <45 DAYS

    Transform the code

    390K LOC to C# .NET Core, 99.96% automated

  3. 03 · DATA

    Schema & ETL

    Tandem schema transformed, validated, loaded

  4. 04 · DAYS, NOT WEEKS

    Frameworks on AWS

    Legacy frameworks rewritten and deployed

  5. 05 · VERIFY

    Refactor & scan

    SonarQube and Checkmarx quality gates

What changed, layer by layer

Layer
Before
After
Language
HP NonStop Tandem COBOL
C# .NET Core
Data
File-based Tandem database; relative and index files
Relational tables in Microsoft SQL Server
Architecture
Monolith: logic and files on one piece of hardware
Multi-tier, distributed, microservices
User interaction
Conversational screens
Pseudo-conversational MVC web, driven by REST
Interfaces
Tandem-specific
REST and a functionally equivalent C# framework
Hosting
HP NonStop mainframe
Containerized on AWS, Well-Architected model

Blueprints before a single line changed

Within the first few days, TSRI generated an Application Blueprint®: full, hyperlinked documentation of the legacy COBOL application, its interfaces, externals and connected utilities.

A matching Transformation Blueprint® showed what the modernized C# .NET Core codebase would look like before transformation began.

application blueprint · includes
Package structureArchitecture metricsNavigation indicesControl flow chartsStructure chartsDead codeSimilarity analysisData element tablesState machine modelsState transition tablesCause-effect graphsCyclomatic complexity
DATA

Data transformed, migration kept in-house

Once Pitney Bowes had assessed the target code and functionality, TSRI transformed the database schema, validated the transformation and performed a custom extract, transform and load (ETL) for the new schema.

Given the confidential nature of customer and production data, Pitney Bowes performed the actual migration itself.

FRAMEWORKS

Frameworks rebuilt on AWS in days

Using the automation in Janus Studio®, TSRI identified each framework in the legacy application, then automatically rewrote, implemented and deployed it on AWS.

With a team experienced in setting up the pipelines and gates for AWS migration, this phase took days rather than weeks.

03 · Architecture

From a monolith to multi-tier, cloud-native on AWS

The legacy application ran its logic and files on a single piece of hardware. The modernized system is distributed across tiers on AWS: application servers talk to the database over the network, and Tandem-specific interfaces are now REST services.

TARGET ARCHITECTURE · AWSSIMPLIFIED
AWS ACCOUNT · VPC
WEB SUBNET
APP SUBNET
PRIVATE DB SUBNET
AWS SERVICES
Web UI
Web API
TSRI API server
.NET Core · Red Hat
TSRI API server
.NET Core · Red Hat
Batch server
HangFire · .NET Core
Amazon EFS
Shared files
Microsoft SQL Server on EC2
Encrypted volume
CloudWatch
Amazon S3
SFTP
Transfer Family
Transit Gateway
Pitney Bowes data center
Other systems
Simplified target architecture, redrawn from the original case study diagram.
01

From conversational screens to the web

Legacy systems request information as it’s needed during an interaction, which maps poorly to the web. TSRI built a pseudo-conversational pattern from industry-standard building blocks: each interaction starts with a REST API call and closes when the application exits.

The result is cleaner, more maintainable and repeatable whenever needed.

02

Locking and performance, solved in the framework

A new locking API uses SQL Server’s native locking while preserving the Tandem behavior the application depends on. To offset the cost of moving indexed file IO onto the network, TSRI minimized database round trips by prefetching records and adding computed columns and multi-column key indices.

All of it lives in the framework and schema, so none of it touched the code generated from COBOL.

04 · No code freeze

The code kept changing. The project didn’t stop.

Pitney Bowes kept maintaining and developing the application throughout. Each updated legacy baseline was absorbed and the latest C# .NET Core iteration regenerated within minutes, because TSRI’s model-based process isn’t a line-by-line conversion.

ORIGINAL SCOPE350,000lines of HP NonStop COBOL at project start
ADDED MID-PROJECT~40,000lines across multiple baseline updates
MODERNIZED390,000lines, regenerated within minutes of each update

05 · Results

Modern, secure and verified, in a matter of months

Working across three time zones (Pacific, UK/Ireland and India), TSRI delivered a containerized, cloud-native Postage Payment application on AWS, with the Tandem database moved to Microsoft SQL Server.

99.96%automated

390,000 lines transformed

All HP NonStop COBOL modernized to C# .NET Core in less than 45 days, without a code freeze.

<8bugs

From thousands of SonarQube findings

Refactoring left fewer than eight bugs, all remedied, and automatically remediated outstanding vulnerabilities.

0findings

Undispositioned medium-or-above issues

No undispositioned medium or above findings on Checkmarx scans of the generated code.

~97%projected

Lower annual TCO

The AWS business case projects costs falling from multiple millions a year to under $30,000 once fully migrated.

CLOUD-NATIVE

Containerized on AWS, Well-Architected, REST interfaces.

KNOWLEDGE TRANSFER

Blueprints® guide current and new developers.

MONTHS, NOT YEARS

Code and data transformed in a matter of months.

Get modern fast.
Stay modern forever.

See Juno in action!