Question on Scalability

[David Thomas]

How does Spark handle the situation where the RDD does not fit into the
memory of all the machines in the cluster together?