Investigation of Power Consumption in Heuristic Cloud Computing Methods

This study investigates the power consumption of various heuristic methods for task scheduling and resource allocation in cloud computing environments. Cloud data centers consume enormous amounts of electrical energy, leading to high operational costs, increased carbon footprint, and reduced hardware lifespan. Heuristic methods (such as Genetic Algorithm, Particle Swarm Optimization, Simulated Annealing, Ant Colony Optimization, and Harmony Search) have been widely proposed as efficient solutions for optimizing task scheduling, load balancing, and resource allocation with the goal of minimizing power consumption while maintaining performance (Quality of Service). However, a systematic comparison of these methods from the perspective of energy efficiency is lacking. This research employs a simulation-based comparative approach using the CloudSim toolkit with ۱۵ different workload scenarios (ranging from ۵۰۰ to ۵۰۰۰ virtual machine requests) and three data center configurations (small, medium, large). Five well-known heuristic algorithms (GA, PSO, SA, ACO, and HS) were implemented and evaluated against a baseline (Round Robin) and an optimal bound (theoretical minimum). Metrics measured included total energy consumption (kWh), Makespan (total completion time), Service Level Agreement violations, and energy-performance trade-off ratio. Findings from the tables showed that the Hybrid Genetic-Particle Swarm Optimization (Hybrid GA-PSO) method outperformed all other single heuristic methods, achieving an average reduction of ۳۸% in energy consumption compared to Round Robin (p < ۰.۰۰۱, effect size ۱.۸۵). Among single heuristics, Simulated Annealing showed the best energy efficiency in small-scale scenarios (up to ۱۵۰۰ VMs), while Genetic Algorithm performed better in large-scale scenarios (above ۳۰۰۰ VMs). Ant Colony Optimization had the highest computational overhead, making it unsuitable for real-time applications. Harmony Search showed moderate performance with high stability. The discussion focuses on the trade-off between energy consumption and makespan, the impact of workload characteristics (CPU-intensive vs. I/O-intensive) on algorithm selection, and the necessity of hybrid approaches. Recommendations include developing adaptive hybrid heuristics that automatically select or combine algorithms based on real-time workload conditions, and integrating machine learning models to predict optimal scheduling parameters.

کلیدواژه ها:

Cloud Computing ، Power Consumption ، Heuristic Algorithms ، Task Scheduling ، Energy Efficiency.