A SYSTEMATIC REVIEW OF BIO-INSPIRED AND MACHINE LEARNING-BASED MULTI-TASK SCHEDULING TECHNIQUES IN CLOUD COMPUTING ENVIRONMENTS

Abstract

In the domain of Cloud Computing multi-task scheduling has many various computational complexities, thus it is classified as a very complicated and computationally hard task due to all the different aspects of cloud computing; e.g. heterogeneity, dynamic and very distributed environments, and large volumes of overall resources. This paper reviews the bio inspired/metaheuristic/machine learning (ML) based approaches that are used to improve the performance of task scheduling in cloud systems. It classifies the majority of existing solutions into three main paradigms; bio inspired optimization methods, ML driven scheduling algorithms, and hybrid intelligent systems. The authors conduct an evaluation of different bio-inspired algorithms like Genetic Algorithms/Ant Colony Optimization/Particle Swarm Optimization and Artificial Bee Colony algorithms on their optimization capability and their performance metrics (i.e., makespan / energy consumption / load balancing and resource utilization). Further the authors evaluate a number of different types of ML techniques such as clustering, regression and reinforcement learning; which allow for adaptive scheduling predictions as well as enabling adaptive decisions based on resource availability. Additionally, different hybrid approaches where metaheuristic algorithms are combined with ML algorithms are carefully evaluated as these types of systems have shown improvements in overall scheduling efficiency and adaptability. The authors conclude the review of the current state of the art in intelligent scheduling in cloud computing by identifying areas that present key research challenges (e.g., scale, real-time decision making and lack of standard benchmarking datasets) and by providing future directions for advancing the research surrounding intelligent scheduling in Cloud Computing environments.