Mobile data access is suffering the curse of the computationally enhanced increase of smart phones, which overloads the traditional cellular network. This paper tries to offload the cellular network traffic through Delay Tolerant Networks (DTN) formed by the short-range communication technologies in these smart phones (e.g., WiFi, Bluetooth). In particular, the content providers deliver L initial replicas of the same data through cellular network to L users. Then those users further propagate the information among all the subscribed users through opportunistic communications. However, phones are carried by different users who may be selfish and not want to help others. If the initial replicas are delivered to these selfish nodes, other users may not get data timely. This paper proposes a method to select the proper number of initial replicas and an optimal policy using the Markov process to distribute the limited initial replicas. Simulation results show the accuracy of the model, and numerical results show that our optimal policy can get better performance.