“Many analytics computations are dominated by iterative processing stages, executed until a convergence condition is met. To accelerate such workloads while keeping up with the exponential growth of data and the slow scaling of DRAM capacity, Spark employs...
