Spark is faster in iterative operations as it supports in-memory computation which is the major drawback of hadoop(map-reduce). In map-reduce intermediate results of each iteration need to be stored on disk and read again from disk in next iteration whereas in spark it is stored and retrieve from memory only.