A new model of cluster computing has become widely

popular, in which data-parallel computations are executed

on clusters of unreliable machines by systems that automatically

provide locality-aware scheduling, fault tolerance,

and load balancing