Databricks cluster policies restrict the configuration options available when users create compute. They can limit node types, worker counts, runtime versions, and idle-time settings.
Policies reduce cost variance by preventing oversized or unbounded clusters. They do not reduce the DBU rate of a permitted configuration. The policy must match the workload; a limit that is too low can increase runtime and total cost.
Use policy rules for interactive and scheduled workloads separately. Review policy usage against DBUs, cloud infrastructure, runtime, and queue time. Databricks documents cluster policies.