We consider the planning problem of a unique cellular model in which parked vehicles, termed Vehicular Relay Nodes (VeRNs), assist the network by relaying data to and from users. VeRNs can lower cellular network load and provide a scalable solution to the explosive growth in data demand. The VeRN model presents a new paradigm that significantly differs from toady's cellular deployments that are based solely on fixed infrastructure. Consequently, known deployment planning methods are inapplicable. The complexity arises from utilizing vehicles which adds stochastic resource variables to the stochastic data demand variables, as well as from the deployment structure/topology and the cost trade-offs between fixed resources, stochastic resources, and user demand. We present a holistic solution for this unique deployment planning problem. Our solution is based on the combination of three methodologies, which enables us establishing the convexity, and thus greediness properties of the problem. This allows focusing the analysis on deriving marginal values, instead of full system analysis. We further propose an approach to approximate the marginal values supported by effective state space bounds. Numerical results show that the method is efficient, and addresses various distribution settings.