A mixed queueing network model of mobility in a campus wireless network

Although wireless networks have become ubiquitous, surprisingly few models of user-level mobility have been developed and validated against traces of measured user behavior. In this paper, we develop and validate a simple mixed queueing network model of user mobility among access points in a campus network. We identify two classes of users, an open and a closed class, corresponding to mobile users that visit the network for a short time before departure, and users that are always resident in the network during the observation period. Using CRAWDAD traces of user-access-point affiliation over time, we compare model-predicted performance with the performance actually observed in the traces, and find that such a mixed queueing model can indeed be used to accurately predict a number of performance measures of interest.