Hadoop Benchmark - Input splits are first sampled - View it on GitHub