ContentsBenchmark method & apparatusBenchmark resultsDebugging the benchmarkDomain experts weigh inExamining the loadThe “right” loadA “fair” loadDo these improvements apply to the other cases too?…but… why is the success rate still weird?Examining the benchmark toolCharacterising the failure mode(s)Overlooked cluesA misunderstood workaroundIt’s never the compiler or the kernel… except when it isConclusionA few weeks ago, Axel Roest published a simple web server comparison, that turned out to no...