Transitioning from high-level business process modeling to physical database implementation frequently introduces structural inconsistencies such as data redundancy and referential integrity violations. To mitigate these issues in modern cloud-native environments, this study proposes a staged transformation methodology. By formalizing the mapping between UML class structures and relational database entities, the methodology integrates web-specific constraints at the conceptual phase: stateless interaction, optimistic locking, and soft deletion. The rules are applied to three reference models of increasing structural complexity, from a three-class e-commerce subsystem to a twelve-class logistics telemetry schema, and evaluated against the classical mapping through an explicitly stated analytical model. Across the three scenarios the classical rules leave between 12 and 51 schema-level design decisions undetermined, while the proposed rules resolve all of them; identifier collisions and silent overwrites fall to zero, and historical retention after 450 delete cycles rises from between 16% and 51% to complete. The cost of these guarantees is quantified as well: the constraints add 37 to 39 bytes to every generated row, between 28.9% and 54.2% of the row width. The approach offers a more robust foundation for scalable web information systems and identifies where its overhead becomes the dominant consideration.

Unified Modeling Language (UML) Relational Database (SQL) Model-Driven Architecture (MDA) Web Information Systems Cloud-Native Constraints Stateless Interaction Data Integrity