Using more than one output group in a Router transformation might cause poor performance. To avoid performance impact, configure the following properties:

Property and Value	Description
infaspark.sql.forcePersist=true	Persists data in the memory and disk to avoid repeated reads for each output group.
spark.rdd.compress=true	Enables compression to store data more efficiently.

You can set these properties either in the Spark advanced properties for the Hadoop connection or at the mapping level. If you set the properties in the Spark configuration, the values apply to all mappings that you run on the Spark engine.