Modeling of congestion-related phenomena has traditionally depended on some very simple descriptions of the way in which traffic is generated. In fact, a great deal of the queueing literature has focussed on models in which the arriving traffic is assumed to follow a Poisson process. However, a large number of recently studied applications appear to demand non-traditional traffic models in order to generate performance predictions that correspond to either what is observed in reality or expected from a prospective technology. In this talk, we will describe several applications and modeling contexts in which non-Poisson traffic flows appear naturally and we will discuss their corresponding impact on performance predictions for queueing systems.