At a high level, accumulators and broadcast variables both are Spark-shared variables. In distributed computing, understanding closure is very important. Often, it creates confusion among programmers in understanding the scope and life cycle of variables and methods while executing code in a cluster. Most of the time, you will end up getting :
org.apache.spark.SparkException: Job aborted due to stage failure: Task not serializable: java.io.NotSerializableException: ...