What is the key improvement of Spark over MapReduce?

Question $4$ options:

$1)$

In memory processing.

$2)$

RDD

$3)$

Distributed computing

$4)$

fault tolerance