Why does calling cache take a long time on a Spark Dataset?

Question

I\'m loading large datasets and then caching them for reference throughout my code. The code looks something like this:

val conversations = sqlContext.read
  .fo