It proved difficult to find samples on how to save files to HDFS in Parquet format when you are not using DataFrames - in other words, you are using RDDs directly. So here is a sample on how to read and write both RDD and DStreams.