Part I — What you are sizing

Which quantities size a system, where they come from, and how much any of them is worth?

Four chapters. The first is about what a single number is worth, and it is the reason for the other three. Those three begin the model the book carries all the way through: a web service and its data on a fleet of hosts, sized in Part III and costed in Part V. All of them are about demand: what arrives, what accumulates, and how sure anybody is of either.

ch01 · Point estimates is the argument the rest of the book answers. A point estimate is not wrong. It is silent, about the spread it threw away at the first multiplication and about the kind of error no amount of sampling can see. It also draws the line this book is built on, between a model whose structure nobody doubts and one that rests on a measured constant or runs into a limit.

ch02 · What a workload is separates rates from levels, and what the world does to you from what you have decided to do about it. It writes the first nodes of the web service model. It ends with a file that runs and a table that cannot yet tell the two apart.

ch03 · Where the numbers come from is the difference between a number you measured, a number you were told and a number you decided. That difference is invisible once all three are cells in one column. It is also where a measured constant stops being a fact about the world and becomes a fact about some software at some version.

ch04 · Peak, mean and growth picks the number in a demand curve that sizes you, and then turns to growth. Growth is the input that moves the answer most in this book, and the one no amount of provenance discipline can turn into a measurement.

Nothing here is about what a system does with the demand. That is Part II, and it is where the arithmetic stops being multiplication.