DYNAMIC RESOURCE ALLOCATION IN CLOUD ENVIRONMENTS USING DEEP REINFORCEMENT LEARNING

Abstract

Cloud environments face continuous fluctuations in workload demands, requiring efficient resource allocation mechanisms to prevent performance degradation and reduce operational costs. Traditional rule-based and heuristic approaches struggle to adapt to dynamic and large-scale cloud infrastructures. This paper proposes a Deep Reinforcement Learning (DRL) framework for autonomous and optimized resource allocation in cloud systems. The model continuously observes workload patterns, VM utilization, and QoS constraints, adjusting allocation policies in real time. Experimental evaluation demonstrates significant improvements in task throughput, reduced latency, and lower energy consumption. The proposed solution offers a scalable, intelligent, and self-adaptive resource management strategy for modern cloud computing platforms.