To solve the problem that the performance of failure detection is effected by mobility of vehicular cloud computing, a new failure detector (VC-FD) based on architecture of vehicular cloud computing was proposed. We found that vehicular cloud computing has a hierarchical structure and the vehicles can be grouped through Roadside Unit (RSU), according to the architecture analysis of vehicular cloud computing. On this basis, the sharing mechanism of detection results is introduced, and the new failure detector VC-FD for vehicular cloud computing was proposed. Then, the model of cooperative failure detection was proposed to evaluate the detection time, mistake rate and detection overhead. The performance of the VC-FD and non shared failure detector was compared in a simulation environment. The experimental results show that the VC-FD can ensure the accuracy, and significantly improve the node failure detection time by increasing limited detection overhead. Thus, the VC-FD can accurately and quickly find out the node failures in vehicular cloud computing, and effectively reduce the influence of vehicular mobility on the performance of failure detection.