Critical Evaluation Of Scheduling And Ranking Algorithms In Fog-Cloud Computing

Abstract

The rapid growth of fog and cloud computing has created a need for the development and implementation of intelligent scheduling and ranking algorithms to improve the allocation and execution efficiency. In the study, the efficiency and evaluation of different heuristic, metaheuristic, and artificial intelligence-based scheduling and ranking algorithms, such as First Come First Serve (FCFS), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Deep Reinforcement Learning (DRL), in the context of fog and cloud computing environments, have been explored. The experimental study reveals that the proposed DRL-based approach has better efficiency and the minimum makespan, i.e., 61.3 seconds, for the execution of 500 tasks, compared to the traditional FCFS approach, which has a makespan of 84.5 seconds. In terms of energy consumption, the proposed approach has the minimum energy consumption, i.e., 454.6 joules, compared to the traditional FCFS and GA approaches, which have energy consumption of 679.7 joules and 567.7 joules, respectively. The proposed approach has the highest CPU utilization, i.e., 78.6%, and the least idle time, i.e., 12.3%, for the execution of tasks, and the minimum average response time, i.e., 119.5 milliseconds. These results demonstrate the limitations of traditional approaches in dynamic fog-cloud environments and verify the potential benefits of adaptive AI-driven models in enhancing efficiency in scheduling, reducing latency in scheduling, and saving power in scheduling. These results provide a foundation for the design of sophisticated scheduling strategies to meet the needs of real-time distributed computing environments. These results demonstrate the limitations of traditional approaches in the dynamic fog-cloud environment and verify the potential benefits of the proposed adaptive AI-driven models to efficiently tackle the problem of scheduling efficiency, latency reduction, and power saving. These results provide a foundation for the design of sophisticated scheduling strategies to meet the needs of real-time distributed computing environments.