GeeCON 2024: Jarosław Pałka, Michał Jonko - Testy wydajnościowe i inne kłamstwa

It is simply a well-known fact that fewer of us care about the performance of our code, even less of us have dealt with performance tests. Among those of us who have an unequal conflict with efficiency, a tiny fistful of us are aware of how many lies, understatements, and false promises lie hidden in them. During the presentation, we will learn anti-patterns in performance investigating and any battle-tested applicable advice on how not to be fooled by test results. Should we be afraid of 'coordinated omission'? Why statistic is the highest form of lying, aka a fast course in reading test results. Why are 'latency' and 'throughput' not the same? Why are test datasets as crucial as test scenarios? What questions do microbenchmarks answer? Do performance tests have layers? Why isn't performance investigating about numbers? These are just any of the topics I will talk about. If you are not curious in performance investigating and avoid this subject like the plague, you will find out why in this presentation. However, if you feel like a 'higher priest of the machine' and you are acquainted with performance topics, I will aid you plan tests with greater awareness and interpret their results.