Cost is one thing we didn't really get to grips with. They use a fairly abstract Databricks Processing Unit - https://databricks.com/product/aws-pricing https:

Cost is one thing we didn't really get to grips with. They use a fairly abstract Databricks Processing Unit - https://databricks.com/product/aws-pricing https://databricks.com/product/aws-pricing - and the costs felt disjointed compared to the workloads we were running. This said, cost didn't really spiral or become an issue for us, especially when you take into account the cost avoidance of administering the Spark cluster and all of the tools. However, the pricing model did feel a little opaque.