The high mobility of VANET makes information exchange across the network excessively difficult. Traditional approaches designed for stationary networks are not applicable due to the high dynamics among the nodes. Applying the routing techniques tailored for general mobile networks inevitably brings huge traffic burden to the crowded urban VANET and leads to low efficiency. To make the information exchange fluent and efficient, we explore the unique features of the urban VANET. By exploring the invariants in the mobile network topology, we are able to efficiently manage the information on top of the "intersection graph" transformed from the underlying network of road segments in the urban area. Our approach can thus achieve efficient query dissemination and data retrieval on this information organization. We intensively investigate and analyze a trace that records the movement of more than 4000 taxies in the urban area of Shanghai City over several months. We grasp the key impact of the fundamental factors that affect the VANET behaviors and accordingly develop tailored techniques to maximize the performance of this design. Experimental results validate the effectiveness and efficiency of our design.