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: When creating a SQL warehouse using the UI or the API and serverless is not enabled, 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