Scholarlink

<p>Virtual Network Embedding can allocate physical resources through network virtual nodes to maximize resource utilization. However, this technology still faces problems such as poor response performance and low average revenue. Therefore, this article innovatively proposes a virtual network embedding model that simultaneously considers service quality requirements and differentiated services. This model combines deficit weighted cyclic algorithm and differential evolution algorithm to synergistically optimize service quality requirements and differentiated services. We have also developed an intelligent mapping mechanism based on graph convolutional networks to achieve efficient and accurate mapping from virtual networks to physical infrastructure. During algorithm iteration testing, the accuracy reaches 95.1%, and the loss rate is only 1.0%. The request acceptance rate of the proposed virtual network embedding model remains stable at around 90%, with an average node and link resource utilization rate of 95.1 and 97.4%, respectively. All performance indicators are significantly better than the comparison model. The results indicate that the proposed virtual network embedding model successfully integrates resource scheduling optimization with deep learning of graph neural networks, effectively solving the performance bottleneck in traditional virtual network embedding techniques and providing innovative solutions for efficient management and personalized services of future network resources.</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

QoS- and DiffServ-aware virtual network embedding using GNN and evolutionary algorithms

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月4日

查看全文
  • Guiyong Sheng,
  • Li Zhang,
  • Jun Pan,
  • Fujun Wang

摘要

Virtual Network Embedding can allocate physical resources through network virtual nodes to maximize resource utilization. However, this technology still faces problems such as poor response performance and low average revenue. Therefore, this article innovatively proposes a virtual network embedding model that simultaneously considers service quality requirements and differentiated services. This model combines deficit weighted cyclic algorithm and differential evolution algorithm to synergistically optimize service quality requirements and differentiated services. We have also developed an intelligent mapping mechanism based on graph convolutional networks to achieve efficient and accurate mapping from virtual networks to physical infrastructure. During algorithm iteration testing, the accuracy reaches 95.1%, and the loss rate is only 1.0%. The request acceptance rate of the proposed virtual network embedding model remains stable at around 90%, with an average node and link resource utilization rate of 95.1 and 97.4%, respectively. All performance indicators are significantly better than the comparison model. The results indicate that the proposed virtual network embedding model successfully integrates resource scheduling optimization with deep learning of graph neural networks, effectively solving the performance bottleneck in traditional virtual network embedding techniques and providing innovative solutions for efficient management and personalized services of future network resources.