FogWeaver: A Multi-Objective Optimization Strategy and Characterization of a Hybrid Internet of Things (IoT) Environment

Abstract

As the complexity and requirements of Internet of Things (IoT) systems evolve, there comes a need for adaptable optimization strategies that improve the process of task allocation, particularly across hybrid fog infrastructures. In this thesis, we introduce FogWeaver, an optimization strategy for efficiently allocating tasks across hybrid fog environments. FogWeaver employs a multi-objective optimization technique to create a metric of comparison between fog- and cloud-based environments while adapting to the needs of an IoT application. By identifying application requirements, it is then possible to determine the necessary resources required for completing computational tasks. To this extent, FogWeaver is designed such that it considers all available computational nodes that exist across both fog and cloud environments making it suitable for IoT applications that can operate within hybrid environments. In addition, FogWeaver employs an adaptable scoring method based on the particle swarm optimization technique to identify an optimal strategy for the allocation of fog- and cloud-based computational resources. Optimizing the allocation of resources helps reduce the costs associated with the execution of tasks or operations across fog environments. In addition, there may exist a number of distributed nodes across a fog environment that can perform a particular task. Hence, as part of our optimization strategy, we attempt to identify a number of computational nodes that can form a supernode that can perform a particular task. By considering the computational capabilities of fog nodes, it is then possible to enhance the resilience and scalability of IoT applications while minimizing the costs associated with running advanced computational tasks such as Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), among many others