Cloud Computing at the Extreme Edge - A Review of Mist Computing and an Evaluation of its Practical Applicability in Object Detection Systems

In today's world of digitalization, data is collected in every conceivable environment. Public transport on highways and roads is no exception. In this area, for example, data is collected about the flow of traffic, cars entering and leaving cities, and information about the presence of vignettes or paid tolls. Often, such traffic monitoring applications are based on camera systems that record data from their environment and send it to the background infrastructure, which processes it using object detection algorithms. This kind of system architecture is also the focus of this thesis, which is concerned with evaluating the impact of using mist computing in such systems. The goal is to evaluate in which areas the concept of mist computing can improve the efficiency of object detection-based traffic video systems in contrast to other Cloud-based architectures. For this purpose, a Cloud-only, a Cloud-Fog and a Cloud-Fog-Mist architecture are implemented in this thesis. These are evaluated and compared with respect to various performance metrics. The result of this work shows that the concept of mist computing can be successfully used to reduce both the throughput through the network and the number of frames sent in the evaluated test environment. Using the applied test methods, a reduction of the network throughput by up to 40% and a decimation of the sent frames via the camera system by up to 39% was measured.