In the traditionalstudy of mobile cellular systems, all users are assumed to havethe same behaviour: they have the same probability ofmaking/receiving a call, and they will move around the networkwith identical mobility. Moreover, the underlying user network isassumed to be fully connected, implying that all users know eachother and they will communicate with each other with the sameprobability. In a practical environment, each user has a differentlist of acquaintances including relatives, friends and colleagues,with whom the user is most likely to make contact, and the size ofthe list varies with individual users. In addition, differentusers will acquire different levels of mobility, depending onvarious factors such as job or age. To evaluate the performance ofa mobile cellular system more realistically, we model the usernetwork with a scale-free network in which the number ofacquaintances of a user follows a power-law distribution.Moreover, the mobile system is studied under three user mobilitydistributions: zero mobility, identical mobility and power-lawmobility. Finally, the system performance indicators, such as thecarried traffic and the blocking probabilities, are evaluatedbased on computer simulations, which show that both the usernetwork configuration and the mobility of users affect theperformance of the mobile system.