In vehicular cloud computing (VCC) systems, the vehicle cloud (VC) consists of computing resources for multiple vehicles to assist the remote cloud (RC) in making real-time decisions on driving vehicles, to improve road safety and driving comfort. The acquisition of additional cloud facilities to increase the capacity of VCC systems when processing intensive requests for cloud resources is costly. A recent approach uses multiple remote microclouds (RMCs) to assist RCs in processing requests because they are smaller and less costly than RCs. Therefore, this paper presents a VCC system with multiple RCs and RMCs, in which an algorithm for allocating request resources among VCs, RMCs, and RCs. Subsequently, this paper proposes an improved simulated annealing algorithm (ImSA) based on density of vehicles in a geographical area to deploy RCs and RMCs with the minimal deployment cost and response cost. For real applications, the accuracy and effectiveness of the proposed algorithm are evaluated on the basis of the real traffic data from three different areas (i.e., urban area, outskirts, and suburban area) in one week.