Vehicular networks are expected to accommodate a large number of data-heavy mobile devices and multiapplication services, whereas it faces a significant challenge when we need to deal with the ever-increasing demand of mobile traffic. In this paper, we present a new paradigm of fifth-generation (5G)-enabled vehicular networks to improve network capacity and system computing capability. We extend the original cloud radio access network (C-RAN) to integrate local cloud services to provide a low-cost, scalable, self-organizing, and effective solution. The new C-RAN is named enhanced C-RAN (EC-RAN). Cloudlets in EC-RAN are geographically distributed for local services. Furthermore, device-to-device (D2D) and heterogeneous networks are essential technologies in 5G systems. They can greatly improve spectrum efficiency and support large-scale live video streaming in short-distance communications. We exploit matrix game theoretical approach to operate the cloudlet resource management and allocation. A Nash equilibrium solution can be obtained by a Karush-Kuhn-Tucker (KKT) nonlinear complementarity approach. Illustrative results indicate that the proposed resource-sharing scheme with the geodistributed cloudlets can improve resource utilization and reduce system power consumption. Moreover, with the integration of a software-defined network architecture, a vehicular network can easily reach a globally optimal solution.