Building Event-Driven Microservices, 2nd Edition

Chapter 7. Denormalization and Eventification

Note

Portions of this chapter previously appeared in Building an Event-Driven Data Mesh (O’Reilly, 2023).

Events are often sourced from normalized data models, particularly if you’re using change-data capture (see Chapter 6) to load in data from relational tables. Relational databases resolve relationships at query time, and are built on the premise that they have access to all necessary data at query time.

In contrast, event-driven microservices rely on data provided by events, and tend to work best when the necessary data is all within the event itself. The reality is that the semantics of streaming joins are far more complex than those over a static data set. While there are streaming frameworks that can resolve streaming joins, they tend to be slower and more expensive than those for static data sets. Discussions of the streaming frameworks that support streaming joins are covered in more detail in Chapters 12, 13, and 14.

Eventification is an implementation pattern for denormalizing and remodeling highly relational streams into forms that are more suitable for event-driven consumers. The end goal is to reduce the amount of repetitive, error-prone, and expensive denormalizations that each consumer must do, when sourcing data from relational systems. It is more challenging to deal with denormalization in a set of event streams than it is a set of tables, so you must carefully consider what should and what should not go into a denormalized ...