| Abstract |
'Smart Anything' needs computation to be performed close to the device. This concept is realized by fog computing, which has applications in various domains such as smart city, smart healthcare, autonomous vehicles, etc. The realization of these applications is contingent upon the availability of physical fog nodes in the vicinity which can be leveraged dynamically by fog tenants in a self-service manner to deploy such applications. Considering the dynamic and elastic nature of application deployment requests as well as the mobility of end-users, the infrastructure resource requirements at a given location by a specific application service offering are unpredictable at best. This necessitates the availability of not just individual fog nodes, but a fog computing environment that provides the features offered by a cloud computing environment such as self-service, availability, elasticity, scalability, performance, etc. To this end, we have proposed the logical organization of fog nodes hierarchically, referred to as Hierarchical and Autonomous Fog Architecture (HAFA). Leveraging such logical organization, we have proposed a scalable, low-overhead, and fully distributed approach to select a cost-efficient fog node, considering both computation and communication costs, from the set of prospective fog nodes to host the given application service. The approach does not require the knowledge of the complete system state by any system entity. We have implemented the solution in a simulation environment and compared its performance with several approaches along with a centralized approach. |