Skip to main content

Dedicated compute overview

This page provides an overview of dedicated compute access mode.

What is dedicated compute?

Dedicated compute is compute configured with dedicated access mode. This means the compute resource can only be used by the single user or group assigned to it. For information on configuring access modes, see Access modes.

Access mode selection

Access mode is configured when creating an all-purpose or job compute resource. The access mode setting is under the Advanced section in the compute UI and represented by data_security_mode in the API.

By default in the UI, access mode is set to Auto, which means the access mode is automatically chosen for you based on your selected Databricks Runtime. Auto defaults to Standard unless a machine learning runtime or a Databricks Runtimes lower than 14.3 is selected, in which case Dedicated is used.

When to use dedicated compute

Choose dedicated compute for workloads that require:

  • RDD APIs: Workloads that need direct access to Spark's RDD (Resilient Distributed Dataset) APIs
  • GPU instances: Workloads that require access to GPU resources
  • R language support: Data science workloads that require R programming capabilities
  • Privileged machine access: Applications requiring access to lower-level system resources
  • Custom containers: Containerized environments and custom software stacks

For most other workloads, including general data engineering, SQL analytics, and collaborative data science, use standard compute for better cost efficiency and simplified management. See Standard compute overview.

User and group assignment

  • Single user: Assign compute to one specific user for exclusive access.
  • Group access: Assign compute to a single group for shared access within the group (Public Preview). See Dedicated compute group access.

Limitations and considerations