The exercise of modeling (i.e., abstracting a real system into some kind of representation) has many advantages. First, several properties of the real system can be elicited in the process of building a model. For example, sources of contention and the nature of the workload are better understood when a model is built. Second, a model is a useful guide on what type of measurements to take and what kind of data to collect. One should restrict the data collection effort to the data necessary to obtain input parameters for the model and to validate the model against the real system. Third, a number of interesting metrics can be readily computed from the input parameters even before the model is solved. For example, as it will be shown later in the book, one can obtain bounds on throughput and response time from service demands. Lastly, a model can be used to answer what-if questions about a real system, avoiding costly and time-consuming experiments.
The following chapters of this book provide the quantitative aspects necessary to use queuing models for capacity planning and performance prediction. Most of the models are implemented in MS Excel workbooks, which makes them easy to use.