Parking problem has been considered the biggest trouble for drivers all over the world. Even some smart-parking apps is designed for providing available parking information, it still cannot avoid the conflict issue, that is, several drivers are chasing one parking place. Hence, our work is to propose a smart on-street parking assistance system in order to minimize parking search time. Our work includes three tiers: first, we propose a distributed parking assistance system which provides real-time available parking to local drivers from grounded sensor networks. Second, we optimize the network performance in order to reach the service delay constraint. Third, crowdsourced parking information can be pushed to distributed systems for improving information quality.