An Investigation into the Performance Optimization of Cloud Computing Systems using Machine Learning Algorithms
Sujon Sarkar · SSRN Electronic Journal · 2025
Cloud computing is now the very backbone through which modern digital services operate, allowing on-demand access to computational resources and storage in an elastic manner. Nevertheless, the complexity, heterogeneity, and dynamic workload patterns of cloud environments have led to extensive challenges in performance management. Traditional static or rule-based optimization techniques that are often used are generally unable to accommodate real-time demand changes. Some performance issues resulting from this include inefficient resource utilization, service degradation, and increased operational costs. The objective of this study is the fusion of machine learning (ML) algorithms into improving the overall performance and efficiency of cloud computing systems. Some of the key performance indicators used in this study include CPU and Memory Utilization, latency, throughput, network traffic, and energy consumption. A comparative analysis has been carried out using the different ML paradigms, including supervised learning (for example, regression, decision trees), unsupervised learning (for example, clustering, anomaly detection), and reinforcement overhead and discusses what this could mean for the practical deployment of ML in production-grade cloud systems. Overall, this investigation confirms that intelligent, autonomous cloud operations can be achieved through ML-based optimization strategies far exceeding traditional heuristics. Hence, it opens a path toward creating self-optimizing cloud platforms satisfying the requirements of increasingly complex, real-time, applications.