I read on http://spark.apache.org/ that Spark supports cyclic data flow.
How can data flow be cyclic if the RDDs used by Spark are immutable?
I thought the processing done by Spark looks like an acyclic graph where the nodes are RDDs and the edges are transformations.
