Human Mobility in Virtual and Real Worlds

The performance of networked systems that support common human activities, such as mobile communication and participation in online games, depends on the characteristics of human mobility. Human mobility models are crucial in the design and tuning of these systems, but depend on detailed and long-term human mobility traces, which, in turn, are difficult to obtain from real environments. Thus, although many human mobility models have already been developed, few have been validated against adequate traces. In contrast, in this work we investigate the use human mobility traces collected from realistic, human-like virtual worlds for the purpose of understanding human mobility and developing new mobility models. We collect traces for over 30,000 virtual citizens of a popular virtual world, and compare these traces with other realand virtual-world traces. Our analysis reveals that the mobility traces of humans in virtual and real worlds have many similar characteristics. We further propose SAMOVAR, a model for human mobility in virtual and real worlds that takes into account mobility characteristics such as the population-wide and individual area popularity. Based on synthetic traces generated from our and three other models, we conduct a simulation-based analysis of the impact of human mobility on the performance of virtual and real networked environments. We find that SAMOVAR leads to useful insights, and discriminates better than the other models among the performance of different networked environments. Wp 1 http://www.pds.ewi.tudelft.nl/∼siqi/ Shen, Brouwers, and Iosup Wp Human Mobility in Virtual and Real WorldsWp PDS