On-demand deployment of containerized Micro-Services on vehicular fogs.

Abstract

With the vehicular manufacturing advancement, real-time vehicular applications require fast processing of the vast amount of generated data by vehicles, thus a maintained service availability and reachability while driving. These applications use sensed data generated by IoT devices on board to support vehicular applications such as, the self-driving cars, real-time traffic signs updates, or even video surveillance and analysis applications. Fog devices are capable of bringing cloud intelligence near the edge, making them a candidate to such requests. However, its location, processing power, and technology used to host and update services affect its availability and performance while considering the mobility patterns of vehicles. Contemporary work in the literature examines the use of virtual machines (VM) to host the essential services on Road Side Units (RSU). The RSU usage raises many limitations including the difficulty of updating services hosted in VMs, RSUs range of coverage, and other handover problems when considering SDN controller to route traffic between RSUs. On the other hand, the evolvement of the On-Boarding Units helps in performing some of the required processing locally. However, one OBU is still not enough to perform real-time processing of generated data and to enable efficient decision making in critical applications like self-driving cars. In this thesis, we overcome the mentioned limitations by introducing a Kubeadm OBUs clustering technique to enable on-demand micro-services deployment with the least costs possible. Docker Containerization technology is adapted to offer light service installment and smooth services updates based on the application's needs. A hybrid multi-layered networking architecture is proposed to maintain network reachability between the requesting user and available Kubeadm fog cluster. We present a master node election algorithm to select the cluster orchestrator in the most effective way taking into consideration the mobility conditions of vehicles. Cluster failur