If one of your tables is very small, you can do a Broadcast Hash Join to speed up your join. There's notebook on the Databricks Guide on that - search for "BroadcastHashJoin" to find that notebook.

Otherwise, a join operation in Spark SQL does cause a shuffle of your data to have the data transferred over the network, which can be slow. If you query the joined table multiple times, you may consider:

1) Saving your joined table as it's own table.

2) Creating a temporary table that represents the joined table, and then caching that table.

This will prevent needing to shuffle the data and doing the join multiple times.