Useful * as a benchmark * When data is local and doesn't benefit from parallelization * when the bottleneck is memory rather than time