Modernizing Legacy Code - While Not Bringing Down Prod
summary
A legacy Java stack with 1.4 million lines of code (Struts/JSP/EJBs on WebLogic and DB2) is modernized incrementally by finding a product 20% traffic seam, routing that segment to a new API that writes a 4,000‑char XML into a temp DB column, and publishing Kafka messages for teams to consume later. That approach shrinks monstrous 900‑line if/else blocks, proves parallel path diff/diff and Mcato‑style probes, and reportedly reduces cost versus wholesale migration—eventually cut 14 teams to 4 while keeping production running.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 2:13Modernize beyond aesthetics
- 7:52Realigning architecture and teams
- 12:27Finding product seams
- 18:22Questioning assumptions
- 27:16Small changes to avoid risk
- 35:21Continuous modernization
- 41:07Strategies vs rewrites or outsourcing
- 45:40Lessons from failures
- 51:20Q&A
related talks
55mUpgrading a Legacy Java EE App with Style
14mDo Legacy Systems dream of Modernization?
59mLiving with your Legacy
58mLegacy modernization made practical: A live coding example of modernization patterns
48mLegacy, for lack of a better word, is good
54mTesting in Legacy: From Rags to Riches
30mModernization: Taming the legacy and keeping the new house clean
33mTo Refactor or Not: How to Systematically Modernize Your Java Applications