What are the SQL warehouse types?

Databricks SQL supports three warehouse types, each with different levels of performance and feature support.

Defining the warehouse types

  • Serverless: This warehouse type supports all features in the pro SQL warehouse type, as well as advanced Databricks SQL performance features. The serverless SQL warehouse type is not enabled by default. To enable the serverless SQL warehouse type, see enable serverless SQL warehouse types.

  • Pro: This warehouse type supports additional Databricks SQL performance features (compared to classic) and supports all Databricks SQL functionality. The pro SQL warehouse type is enabled by default.

  • Classic: This warehouse type provides support for entry level performance features and a limited set of Databricks SQL functionality. The classic SQL warehouse type is enabled by default.

What are the warehouse type defaults?

  • Serverless enabled: When creating a SQL warehouse using the UI or the API and serverless is enabled, the default SQL warehouse type is serverless.

  • Serverless not enabled and using the UI: When creating a SQL warehouse using the UI and serverless is not enabled, the default SQL warehouse type is pro.

  • Serverless not enabled and using the API: When creating a SQL warehouse using the API with default parameters, the default SQL warehouse type is classic.

Note

For API calls, specifying the SQL warehouse type is recommended.

Comparing features of serverless, pro, and classic SQL warehouses

The following table in this article helps you compare features in Databricks SQL across all three warehouse types.

Databricks SQL Feature comparison

Features

Classic

Pro

Serverless

Overview

Self managed, introductory SKU, compute in your account

Self managed, advanced SKU, compute in your account

Fully managed, elastic, best value

Exploratory SQL

SQL Editor

yes

yes

yes

ANSI SQL

yes

yes

yes

Management and governance

Query history and query profile

yes

yes

yes

Data Explorer support for Unity Catalog

yes

yes

yes

Managed data sharing

yes

yes

yes

Connectivity

Partner connect

yes

yes

yes

Performance

Photon engine

yes

yes

yes

Predictive I/O

no

yes

yes

SQL ETL/ELT

Query federation

no

yes

yes

Workflows integration

no

yes

yes

Data science and ML

Geospatial functions

no

yes

yes

Serverless SQL data warehouse

Instant, elastic, fully-managed compute

no

no

yes