A data model describes how data is organized and how related information fits together. In Azure Cosmos DB, modeling affects application performance, scalability, query efficiency, and cost, especially as a dataset and customer base grow. section 11.0 section 11.3
Cosmos DB data modeling emphasizes three practical choices:
- Use schema-less structures when product records need to evolve. Items can have different fields, but application logic must handle missing fields and preserve consistency. section 11.3.1
- Use denormalization when read speed and scalability matter. Embedding related information in one document can avoid joins and reduce multiple queries. section 11.3.1
- Choose partitioning deliberately to distribute data evenly, avoid hot partitions, and improve query performance and scalability. section 11.3 section 11.7
An evolving product record
A product document can store an identifier, type, material, sizes, colors, supplier details, and availability together. Another record for the same product may contain extra ratings or images, illustrating schema flexibility—and the need for code that does not panic when a field is absent.
