Some common misconceptions about performance modeling and validation

Queueing networks and Markov processes etc. are widely used in modeling computer systems and communication networks to study their performance and reliability. To solve a real world problem, the model developed has to be validated through measured data. In this paper, we point out that in validating a model, one has to be very clear about one's claims regarding what has been validated; Too "accurate" results do not imply a correct model and usually indicates a validation problem. We discuss some common misconceptions in performance modeling and validation. We illustrate our points through examples. To capture the main concepts, the problems are simplified in these examples.