We should avoid using groupByKey because of it $...$

A$.$

Always reads the data from HDFS and causes large data transport.

B$.$

Shuffles all the key$-$value pairs data around and generates lots of unnecessary data transport. Also, it may cause memory problems because when grouping the values by key, all the data associated with a single key has to be collected on one worker node.

C$.$

causes lots of communication with master node and it has lots of costs.

D$.$

generates lots tiny small jobs compared to other transformation operations.