Descriptive statistics of performance distributions

Application performance metrics are never described by a single number. They are random values from some distribution. But how do we describe that distribution? How do we consistently calculate the mean and degree of variation? How do we calculate quantiles and plot the density of a distribution? What to do with multimodality and extreme outliers?

There are dozens of different methods for each of these tasks, each with its own characteristics. In this talk, we will dive into the fascinating world of descriptive statistics and discuss the variety of mathematical tools available. And after the talk, you will have a general understanding of the issues in this area and the adequacy of certain approaches, which will protect you from common mistakes when aggregating performance measurements.