The case of the Clumsy kernel
Brendan Gregg · ;login:: the magazine of USENIX & SAGE · 2014
All benchmarks are wrong until proven otherwise. Benchmarking is an amazingly error-prone activity, with results commonly misinterpreted and wrong conclusions drawn. However, every now and then, a benchmark takes me by surprise and not only is correct but also identifies a legitimate issue. This is a story of debugging a benchmark.