Understanding Before Migrating: How We Used Knowledge Graphs to Transform Legacy Systems

Session track: Modern Applications

Session time: 04:00 pm - 04:30 pm UTC November 12

Session description:

Modernizing legacy systems is rarely a technical challenge alone—it is first and foremost a challenge of understanding. We were tasked with transforming a legacy ecosystem consisting of a black-box mainframe system and a customized MOVEX installation. Documentation was incomplete, access to source code was limited, and understanding the landscape required analyzing more than 6,500 attributes, business processes, user interfaces, contracts, and target-system structures. Traditional approaches based on documentation, spreadsheets, SQL analysis, and interviews quickly reached their limits. Some relationships were documented, some were incorrect, and others were missing entirely. To establish a shared understanding, we built a Neo4j-based metadata graph connecting systems, schemas, attributes, processes, interfaces, contracts, and discovered relationships. As understanding evolved, we introduced a second graph: a semantic transformation graph containing transformed business data organized around business concepts rather than technical structures. Together, the graphs enabled discovery of hidden dependencies, modeling of complex transformations, automatic generation and population of migration load templates, and end-to-end traceability from source systems to target-system imports. This session shares practical lessons learned from building metadata and transformation graphs that turned a legacy landscape into an executable transformation model.