Figure 3: Hadoop 1.x relies on hard partitioning of the existing resources in the cluster, which in turn results in less than optimal use of the available capacity. Jobs that could not be distributed by Map() and Reduce() ran accordingly slower.
