Big data processing in fog: smart parking case study

The introduction of Internet of Things (IoT) and its positive effects within city life context will be first seen in the applications that most obviously affect people's lives, such as improving traffic efficiency, reducing time spent in vehicles while travelling around the city and generally mitigating traffic congestion. Vehicle parking and its management represents one of the major issues that directly affects people's time and can have significant financial effects, thus making it directly interesting to both service providers and users. Parking can be more efficient by making it smarter. This can be achieved by extensive use of IoT-based sensing in carparks, then processing and further contextualising the huge amount of generated data for two types of goals: (1) long-term goals of efficient carparks management and (2) short-term goal of helping the drivers by reducing time for finding a suitable carpark. In this paper we propose an IoT-based platform for monitoring carpark occupancy around a city and then doing data analytics near its sources, at the fog level, without streaming and storing all sensing data in the cloud. The data analytics system in our platform uses Hadoop MapReduce and is run on a cluster of commodity computers at each fog computing node. We explore the efficiency and scalability of the approach by performing data analytics tasks related to smart parking on the parking datasets of various sizes collected from a real sensor-based system and by extrapolating it by significant increase in its size.