As one of the core technologies of the internet of things and smart city, internet of vehicles (IOV) has been rapidly developed in recent years. And it has been playing a great role in improving urban traffic conditions. However, with the increasing size of vehicles and the types of applications, some bottlenecks have appeared such as high heterogeneity, poor mobility support and lack of extensibility. So how to develop the new generation of IOV has become one of the most important research topics. In this paper, we apply the software-defined network (SDN) and cloud computing to the heterogeneous IOV to bridge the gaps. The proposed architecture includes four layers, which can provide flexibility, scalability and mobile support which didn’t exist in traditional ways. Moreover, a services preloading scheme based on trajectory prediction is proposed for effective resource management to minimize communication cost and transmission delay. In order to improve the effectiveness of the scheme, the improved threshold probability suffix tree algorithm is designed by training the historical trajectory and making the real-time prediction. We also evaluate the proposed approaches by experiments. The effectiveness and the efficiency are validated by the results.