Spark : running same spark job multiple times, but having huge performance gap

Question

I am running a Scala script which has many Spark actions.

When I run this script multiple times on my cluster, the performance gap for a single job ( foreachPartition