Intelligent Resource Allocation in Cloud Computing Environments: Leveraging Machine Learning for Dynamic Workload Balancing, Cost Efficiency, and Performance Optimization

Abstract

Cost optimization is widely considered a critical aspect of resource management in current cloud computing scenarios to consider service quality, resource utilization and SLAs. The traditional static or the rule based algorithms are not well suited to handle the dynamics, heterogeneity and variability of workloads usually found in cloud data centers of vast sizes. To address these challenges, the current research introduces an intelligent resource allocation framework that uses machine learning. In particular, the workload prediction is achieved with a Random Forest model, while the resource management, scheduling, and auto scaling is done by a Deep Q-Learning agent. The proposed framework was tested with CloudSim Plus simulator and Google Cluster Trace dataset comparing its performance with FCFS, Round Robin and threshold-based methods. It is seen that the resource utilization increases up to 85.3%, SLA violation is reduced to 4.1% and the cost of execution is reduced by 28% which is based on conventional strategies where the throughput of the tasks achieved is 674 tasks per hour. The results demonstrate how machine learning can help build smarter, efficient, and cost-aware cloud applications and infrastructure to support responsive and autonomously optimizing cloud systems.