Multi-Agent Reinforcement Learning for Efficient Cloud Resource Utilization

Mensah, Kwame, Ivanov, Nikolai

Innovative Multidisciplinary Journal of Applied Technology · 2025-yil

Annotatsiya

Cloud computing has revolutionized modern IT infrastructure by offering scalable and on-demand resource provisioning. However, the dynamic nature of cloud workloads presents significant challenges in efficient resource allocation, often leading to underutilization, service delays, and increased operational costs. Traditional load balancing techniques struggle to adapt to real-time workload fluctuations. To address this, Multi-Agent Reinforcement Learning (MARL) has emerged as a powerful approach for optimizing cloud resource management. This study explores the application of MARL-based frameworks to enhance load balancing, resource scheduling, and energy efficiency in cloud environments. We discuss how multiple intelligent agents can independently learn and coordinate decisions to optimize resource allocation across distributed cloud infrastructures. The research delves into model-free and model-based RL algorithms, highlighting the advantages of Deep Q-Networks (DQN), Actor-Critic methods, and Multi-Agent Deep Deterministic Policy Gradient (MADDPG) in dynamically adjusting resource distribution. Key performance metrics such as latency, throughput, energy consumption, and cost reduction are evaluated to compare MARL-based approaches against conventional cloud management techniques. Real-world case studies from leading cloud service providers (AWS, Google Cloud, Microsoft Azure) demonstrate MARL’s scalability, adaptability, and decision-making efficiency in complex cloud environments. Despite its advantages, computational overhead, training time, and real-time adaptability remain challenges in MARL deployment. The study further explores future directions, including the integration of federated learning, edge computing, and secure MARL models to enhance cloud resource management. By leveraging multi-agent reinforcement learning, cloud service providers can achieve dynamic, autonomous, and self-optimizing resource allocation, leading to improved performance, reduced costs, and sustainable cloud operations. This research contributes to advancing intelligent cloud computing by demonstrating MARL’s potential to revolutionize next-generation cloud infrastructures.

Maqola ma’lumotlari
MualliflarMensah, Kwame, Ivanov, Nikolai
JurnalInnovative Multidisciplinary Journal of Applied Technology
Nashr sanasi2025-03-10
Jild3
Son3
Betlar19-35
TilIngliz

Ilmiy soha

Innovative Multidisciplinary Journal of Applied Technology jurnalidan boshqa maqolalar

Innovative Multidisciplinary Journal of Applied Technology — barcha maqolalar