An experience of finding a vacant parking slot can be very stressful in densely populated areas, especially in peak hours. Such parking process takes a long time, wastes significant gasoline and emits extra vehicle exhaust that harms the environment. Smart parking, aimingto assist drivers in finding desirable parking slots more efficiently through Information and Communication Technologies (ICTs) such as Vehicle Ad Hoc Networks (VANETs), has received extensive attention recently. Current VANETsbased parking slot allocations can not provide a fully satisfactory solution, because vehicle communication devices – on-board units (OBS), and roadside units (RSUs) lack computational capabilities to perform humanized and accurate service provisioning, such as real-time parking slots information and probabilistic prediction on future parking slots. Therefore, we in this paper propose a fog computing based smart parking architecture to improve smart parking in real time. Fog nodes deployed at parking lots, cooperating with each other, enable real-time parking slot information provisioning as well as parking requests processing. The cloud center can further enhance smart parking capability by enforcing global optimization on parking requests allocation. The experimental results of our approaches show higher efficiency compared to other parking strategies. The proposed fog computing based smart parking can lower the average parking cost and minimize gasoline wastes and vehicle exhaust emission.