Schedule - PGConf.DE 2025
Normalize or De-normalize? Relational SQL Columns or JSON Document Attributes?
Date: 2025-05-09
Time: 09:00–09:45
Room: Ballsaal 2
Level: Intermediate
Data modeling has progressed beyond traditional SQL Normal Forms and the unstructured storage of early NoSQL systems. Modern applications demand flexible data models that merge normalized entities with semi-structured documents. Contemporary SQL databases now support binary JSON columns and offer APIs similar to MongoDB (such as FerretDB for PostgreSQL). Furthermore, NoSQL databases have evolved to allow normalized collections and provide transactional operations.
However, key questions remain: When should multiple entities be embedded within a single JSONB document? And when is it more efficient to define a relational column instead of a sub-document attribute?
This session will clarify the distinction between physical and logical data modeling. Denormalization can enhance physical storage by placing frequently accessed data together, while normalization is essential for maintaining logical data integrity and preventing issues related to redundancy. By the end of this session, you will understand how to balance these two approaches to create data models that are efficient, maintainable, and adaptable to changing application requirements.