they are read-only shared variables that are cached and distributed to the cluster nodes to avoid shuffle operations. They can be very useful when we have a distributed machine-learning application that needs to use and load a pre-trained model. We broadcast the model as a variable, and ...