Development of Knot DNS authoritative server is supposed to be driven by benchmarks. Generally we want to test all changes that might affect performance and compare them to the previous versions. The basic question is whether we can see any statistically significant improvement, especially in case the changes in measurement results are small and unevenly distributed. To answer that question we use Hotelling's test and occasionally ANOVA and regression analysis to go a bit deeper and provide developers with information they are interested in. Another application of these methods is assessment of DNSSEC performance impact by comparing different DNSSEC algorithms on the same data sets and same servers.
|Please also consider this submission for the NANOG65 DNS track||Yes|