A Brief Review of Population-based Methods for Task Offloading in Cloud-to-Edge Continuum

Task offloading is one of the most important issues in the cloud-to-edge continuum in order to optimise quality of services metrics and maximize the revenue for users and providers. While the literature includes multiple mathematical, machine learning and control theory models each of them have its limitations such as the lack of a mathematical formulation or the requirement of historical training dataset. Population-based methods including swarm intelligence and evolutionary computation algorithms have many advantages that make them prominent meta-heuristic solutions into many combinatorial problems. Even if various population-based methods have been proposed for specific orchestration tasks in the distributed computing field, there is not any paper that presents and experimental compares them. In this paper we provide a comprehensive overview of the most popular population-based methods and discuss their pros and cons for their applicability for the task offloading challenge. We also make experiments with Google Cluster Dataset and conclude that particle swarm optimization significantly surpasses the other methods.