With an ever-increasing number of vehicle services, connected vehicles (CVs) are becoming more software-dependent than ever, which is bringing a tremendous burden on the infrastructure. However, how to efficiently support such a tremendous amount of services is unknown. This paper takes the first step to examine different vehicle architectures. We first illustrate the evolution of automotive software and computing system. Next, we introduce the concept of software-defined vehicle, classify essential CV applications, and list their performance requirements such as communication mode, critical latency, data rate per vehicle, and communication range. Then, we choose over-the-air (OTA) update as our case study, followed by the introduction of four types of CV architectures. Then, we discuss advantages and drawbacks of each type of architecture. Finally, we conclude this article by presenting an edge-based architecture for CVs, named EdgeArC.