Minimizing data transfers and avoiding shuffling helps write spark programs that run in a fast and reliable manner. The various ways in which data transfers can be minimized when working with Apache Spark are :
Using Broadcast Variable : Broadcast variable enhances the efficiency of joins between small and large RDDs.
Using Accumulators : Accumulators help update the values of variables in parallel while executing.
The most common way is to avoid operations ByKey, repartition or any other operations which trigger shuffles.
