Use single node clusters where spark is not being used or for smaller workloads.
Also, use Databricks jobs for multistage jobs and share the cluster across the tasks/notebooks. Job will finish faster and you won’t pay for the startup time.
In Azure, VM reservations can also reduce cost significantly for regular workloads.
These are solid tips.
A few more:
Use single node clusters where spark is not being used or for smaller workloads.
Also, use Databricks jobs for multistage jobs and share the cluster across the tasks/notebooks. Job will finish faster and you won’t pay for the startup time.
In Azure, VM reservations can also reduce cost significantly for regular workloads.