Understanding Indexing in Azure Cosmos DB

Even though Cosmos DB automatically indexes every property by default, understanding how indexing works in Cosmos DB is vital for achieving efficient query performance.

In Azure Cosmos DB, every property in our items are indexed by default. This is fantastic for developers, as this means we don’t have to spend time managing indexing ourselves.

However, there may be times where we do want to customize the indexing policy depending on the requirements of our workloads. The purpose of this article is to show you how indexing works in Azure Cosmos DB, what kinds of indexes there are in Cosmos DB and how we can employ different indexing strategies to optimize performance depending on what we’re trying to achieve.

How indexing works in Cosmos DB

Azure Cosmos DB persists our items within our containers as JSON documents. We can think of these documents as trees and each property in our item as a node within that tree.

Say that I have a document for a customer, and that customer has multiple address associated with them, we could visualize our property paths for this customer like so:

/address/0/addressLine1: "1 Shipping way"
/address/0/addressCity: "Shipville"
/address/1/adressLine1: "2 Billing Road"
/address/1/addressCity: "Billington"

When our customer item gets written to our container in Cosmos DB, each property path and the value of the property is indexed.

Types of Indexes in Cosmos

Cosmos DB supports 3 kinds of indexes:

1. Range Indexes
2. Spatial Indexes
3. Composite Indexes

Range Indexes are used for equality queries (WHERE someProperty = ‘someValue’), range queries (WHERE someProperty > ‘someValue’), ORDER BY queries, JOIN queries and for checking the presence of a property (WHERE IS_DEFINED(someProperty)). We use Range indexes on string or number values.

Spatial Indexes are used for geospatial objects. Currently, Points, LineStrings, Polygons and MultiPolygons are supported. We can use Spatial indexes on GeoJSON objects.