Scholarlink

<p>To tackle the problems of high energy consumption and resource scheduling in campus edge networks under dynamic service environments, this paper introduces a low-power deployment approach combining deep reinforcement learning and graph convolutional network. Under a software-defined network and network function virtualization architecture, this method utilizes graph convolutional network to extract network topology features to perceive the dependencies between virtual network functions and constructs a deep reinforcement learning agent for dynamic deployment decisions. Simultaneously, a load-aware hybrid scheduling mechanism is introduced. During low-load periods, a heuristic algorithm is used for rapid response, while during high-load periods, the deep reinforcement learning agent is activated for hot migration of virtual network functions and global energy efficiency optimization. Simulation results show that in a small-scale simulation environment, the proposed method achieves a request acceptance rate of 91.1% with 1000 service requests, and in a medium-to-large-scale environment, the request acceptance rate is 88.2%. Regarding power consumption, the average power consumption of the proposed low-power deployment method in the two simulation environments is 132.1&#xa0;W and 178.4&#xa0;W, respectively, with load balancing indices of 0.91 and 0.88. The findings demonstrate that the introduced approach can successfully improve service carrying capacity and resource utilization efficiency, reduce system energy consumption and ensure service quality, providing an effective solution for building a green and efficient campus edge network.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Low-power deployment of campus edge networks based on DRL and GCN in SDN/NFV architecture

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月1日

查看全文
  • Zhun Feng

摘要

To tackle the problems of high energy consumption and resource scheduling in campus edge networks under dynamic service environments, this paper introduces a low-power deployment approach combining deep reinforcement learning and graph convolutional network. Under a software-defined network and network function virtualization architecture, this method utilizes graph convolutional network to extract network topology features to perceive the dependencies between virtual network functions and constructs a deep reinforcement learning agent for dynamic deployment decisions. Simultaneously, a load-aware hybrid scheduling mechanism is introduced. During low-load periods, a heuristic algorithm is used for rapid response, while during high-load periods, the deep reinforcement learning agent is activated for hot migration of virtual network functions and global energy efficiency optimization. Simulation results show that in a small-scale simulation environment, the proposed method achieves a request acceptance rate of 91.1% with 1000 service requests, and in a medium-to-large-scale environment, the request acceptance rate is 88.2%. Regarding power consumption, the average power consumption of the proposed low-power deployment method in the two simulation environments is 132.1 W and 178.4 W, respectively, with load balancing indices of 0.91 and 0.88. The findings demonstrate that the introduced approach can successfully improve service carrying capacity and resource utilization efficiency, reduce system energy consumption and ensure service quality, providing an effective solution for building a green and efficient campus edge network.