The easiest optimization is that if one of the datasets is small enough to fit in memory, it should be broadcasted to every compute node. This use case is very common as data needs to be combined with side data, such as a dictionary, all the time.

Mostly joins are slow due to too much data being shuffled over the network. With the Broadcast join, the smaller dataset is copied to all the worker nodes so the original parallelism of the larger DataFrame is maintained.