In the cluster programming model for MapReduce, we don't need to know how many processors or threads are running. Though we are giving up some performance as data needs to be moved back and forth which is costly.