Distributed network management systems in cloud computing environments

Abstract

Deep reinforcement learning is rapidly emerging as a transformative approach for distributed network management in dynamic and heterogeneous cloud environments. This paper presents a novel intelligent framework that embeds advanced DRL agents with hierarchical feature extraction into cloud-based Distributed Network Management Systems, enabling precise and adaptive control over network resources, topologies, cand service isolation. By constructing a high-fidelity simulation environment based on NS-3, the effectiveness of the framework has been fully validated in both threshold driven and static methods. Research shows that DRL-based systems can still maintain throughput stability in the face of topology changes, node failures, tenant traffic fluctuations, or throughput delays. The framework ensures network performance during large-scale failures and rapid expansions thru robustness and scalability analysis. Despite these advantages, computational overhead and policy convergence are issues during large-scale deployment. The research findings indicate that the key to enhancing real-time cloud network intelligence lies in the architecture based on Deep Reinforcement Learning (DRL). These findings provide useful guidelines for future engineering projects to ensure that network management infrastructure possesses autonomy, flexibility, and efficiency.