What Azure Cosmos DB is: a globally distributed NoSQL database

Cosmos DB sits in the non-relational part of the exam. Understanding what makes it different from a relational database is the starting point for deciding when to use it.

Azure Cosmos DB is a fully managed NoSQL database service, offered as platform as a service (PaaS). NoSQL is the collective name for databases that store data in structures other than relational tables, such as documents, key-value pairs, column families and graphs (Chapter 1, Representing and storing data, describes these four types). PaaS means Microsoft runs the underlying infrastructure, including server provisioning, patching, updates and backups, so the customer only looks after the data and the application.

Key characteristics:

• Schema-agnostic. Items in the same container don’t need to share a structure: one item might have five properties and the next fifteen different ones. This suits applications whose data shapes change over time or vary between records.
• Automatic indexing. Cosmos DB creates and maintains indexes on every property of every item by default, so there is no schema to define upfront and no indexes to manage by hand. An index is a structure that lets the database find data without scanning everything.
• Global distribution. Azure regions can be added to an account at any time, and the service replicates data to each one automatically.
• Low latency and elastic scale. Cosmos DB offers single-digit millisecond response times and automatic, instant scalability. Microsoft runs demanding services such as Xbox Live and Microsoft 365 on it.

Two features are worth a sentence each. The change feed tracks changes to a container so that other services, such as Azure Functions, can react to them in event-driven designs. Vector search stores vector embeddings (numeric representations used by AI applications) alongside operational data, for scenarios such as AI agents and retrieval-augmented generation.

Common trap: Assuming Cosmos DB needs a schema and indexes defined before data is loaded, like a relational table - it is schema-agnostic and indexes every property automatically by default, so items with different shapes can share a container.

Get the whole book

This note is one section of Ultra Transcenders DP-900: Microsoft Azure Data Fundamentals, an independent study guide that explains every topic the exam covers by technology, with comparison tables, diagrams and the common traps, plus a glossary linked to Microsoft Learn.

About the book · DP-900 terms in the glossary · All DP-900 study notes