This paper presents the ways of designing simulation models of web servers. At the beginning Queuing Network models are introduced, those models are generally known and often used in the initial phase of research on particular technical solutions. Next, an entirely new approach to the issue discussed is presented - neuro-fuzzy models, thanks to which it is possible to automate the process of designing simulation models. In Section 4 results of comparative tests of these two models are presented. Based on these results it can be concluded that neuro-fuzzy models are accurate and can be used in simulation research.