Deutsche Bank KreditManager: COBOL and JCL to Java and Python on OpenShift
How TSRI modernized 425,761 lines of COBOL and JCL, migrated DB2 and VSAM data to Oracle, and moved KreditManager to Deutsche Bank’s private OpenShift cloud, with little to no business disruption.
Executive summary
Keeping pace with digital-first banking
Deutsche Bank, a global bank present in 58 countries, saw that its legacy systems lacked the agility to compete with mobile finance apps. Its mainframe ran COBOL and JCL with DB2 and VSAM databases, and leadership knew it needed a cloud-enabled, modern architecture.
After a competitive proof of concept, TSRI modernized KreditManager, the application behind all of the bank’s loan, credit and mortgage processing. Using a step-wise method, TSRI migrated 425,761 lines of COBOL and JCL code and comments to a multi-tier environment on Oracle, with little to no business disruption.
01 · Challenge
Reliable mainframes, not enough agility
Deutsche Bank had a relatively reliable mainframe infrastructure that used COBOL and JCL, running DB2 and VSAM databases. But its legacy systems were becoming a liability: they could not keep up with the quickness of a self-contained mortgage-underwriting app, for example. To stay relevant for customers and ongoing market needs, the bank would need to be cloud-enabled with a modern architecture.
After evaluating several leading automated conversion vendors based on code quality, ROI and cost, as well as schedule through a proof of concept, TSRI was selected as our preferred transformation vendor.
02 · Selection
Chosen after a competitive proof of concept
Based on a report by Gartner Research, Deutsche Bank selected TSRI as one of five modernization companies to run a pilot transforming two banking applications: EZB DepotStatistik in Germany and DBSAE in Spain. Both would move from COBOL to Java to run on the bank’s private on-premises cloud.
Deutsche Bank selected TSRI for its automated refactoring capabilities and the quality of the code generated by the proofs of concept. The scope: modernize 425,761 lines of COBOL and JCL to Java and Python respectively, replace the legacy DB2 and VSAM flat-file, sequential databases with Oracle, and run the application on the OpenShift on-premises cloud.
03 · Approach
A step-wise path to the OpenShift cloud
Deutsche Bank wanted to bring smaller applications online first and modernize more over time to avoid long disruptions. TSRI’s dual database input/output (DIO) layer let some legacy functions keep running alongside the transformed applications. TSRI also added a process that compared new and old datasets in real time, so engineers could uncover errors immediately.
- 01 · PILOT
Proof of concept
EZB DepotStatistik and DBSAE to Java
- 02 · FIRST
KreditManager: Beraten
The Consult branch, migrated first
- 03 · NEXT
KreditManager: Entscheiden
The Decision branch, same results
- 04 · ONGOING
Remaining applications
Migrating the rest of the portfolio
What changed, layer by layer
One refactoring rule, about 12,500 hours saved
TSRI used SonarQube to surface maintainability issues and security vulnerabilities, then wrote refactoring rules to remediate them. One issue in the legacy code appeared about 25,000 times. SonarQube estimated 30 minutes to fix each instance; automating the fix saved approximately 12,500 hours of development time.
04 · Challenges
Firsts, adapted and overcome
An enterprise service bus
The DBSAE application included a collection of APIs known as an enterprise service bus, which had not been transformed before on any platform. TSRI tuned Janus Studio® to transform it and reach the required code quality.
Older language versions
Some Deutsche Bank requirements were only compatible with earlier versions of Java and Python, so TSRI re-migrated some of the transformed code.
COBOL pointers
TSRI’s team had never encountered the memory-management tools called pointers. Janus Studio® was adapted to work with them, a normal adaptation for an unprecedented situation.
Onboarding
Banking regulations and company requirements meant onboarding TSRI as a non-German company took longer than expected.
05 · Results
Modernized at 99.7% automation
Over the course of a year, working across four time zones, TSRI modernized and migrated Deutsche Bank’s legacy code and databases. Because the first transformation encountered no errors, TSRI moved on to the next applications, and continues to migrate them with the same step-wise method.
425,761 lines transformed
COBOL and JCL code and comments, to Java and Python.
Relational database
DB2 and VSAM data migrated to a modern Oracle database.
In error remediation
Plus tens of thousands of hours now spent on innovation.
In the first transformation
KreditManager went live without errors.