Analysis of wide area user mobility patterns

In this paper, we present an analysis of user behavior and mobility patterns based on a trace of accesses to a department e-mail server. In contrast to previous studies, we consider a single service and examine how a user community connects to it while moving across a variety of different service providers' wireless and wired networks. By measuring an e-mail service, one that users access often, we were able to monitor a large number of sessions originating from a diverse set of locations. Our contributions include: a unique approach to extracting user mobility information from traces of client application interactions; a novel approach to modeling user behavior and mobility; and a demonstration of how such models can be used to generate synthetic traces. Overall, although some users are highly mobile, we find most users have a low degree of mobility - 70% of users access their e-mail from 2 or fewer unique locations. We also find that our observed session times are longer than those reported by previous mobility studies in wireless networks.