The polic information system is one of the largest information system of the country and has been running for decades. Currently, it consists of more than ten subsystems and provides more than forty kinds of information services for police officers. The performance of the police information system is a critical issue, since it directly impacts the quality of policing information services Thus, it is important to study the police information system. Recently, the distributed data processing technology has become matured and widely used in data processing industries. However, the data distribution strategies of the police information system have not been well studied and have been developed mainly based on conventional wisdom. Conventional wisdom may provide good solutions as long as the number of system variables in small Hwever, as the number of system variables becomes fairly large, it may be more effective to use scientific analysis methods. This research proposes two feasible distributed system architectures for the police information system. Mathematical performance analysis models based on queuing theory are developed to evaluate the proposed architectures. The system availability of the proposed schemes is also studied.