An Optimized IoT Architecture based on Fog Computing with a new Method of Data Transfer Control

Abstract

Over the years, distributed and grid computing paradigms have evolved to cloud computing, which has become a common approach applied in the Internet of Things (IoT). The growing popularity of the cloud computing paradigm lies mainly in the simple management of end devices, uniform access to many services, elasticity of available resources and cost savings. In addition to these advantages, the expansion of IoT devices and the demand for speed and data volume have provided an opportunity for the emergence of new computing paradigms. The fog computing paradigm brings data processing nearer to the end devices while preserving the cloud connection, leading to lower latency, higher efficiency and location awareness.

The overall aim of the dissertation is the design and implementation of an optimised IoT network architecture which adopts the fog computing paradigm. To eliminate the need to build completely new infrastructure, the optimised network architecture is based on LoRaWAN, which has already been deployed at many locations and offers long-distance communication with low-power consumption. This raises several challenges which need to be overcome. For correct functioning of the fog computing paradigm, it was necessary to explore a new method of controlling the data transfer between IoT gateways and the cloud service. The methods explored in this dissertation are both static (based on predefined values) and dynamic (based on machine learning).