Their documentation is pretty ok: Databricks documentation | Databricks on AWS. Note that just because you are on databricks, you don’t get paralellism by default. The common use case is computing in a notebook using single core tools (such as pandas) but leveraging parallelism in hyper-param search. permalink: databricks
Costs
You pay for the clusters (to your regular provider) and, in addition, “as you go” depending on the computing you are doing, to Databricks.