Advanced Virtual Network Embedding: Combining Graph Attention Network and DRL for Optimal Resource Utilization

Network virtualization (NV) increases service development flexibility and efficiency by enabling multiple virtual networks to operate concurrently on a single physical network. Mapping virtual nodes and links onto physical resources is a critical task known as virtual network embedding. Finding the best solution is extremely challenging because it is an NP-hard problem, particularly for link embedding. Optimizing resource usage, reducing request rejections, and enhancing virtualized network performance all depend on how well VNE algorithms work. Our method combines Double-DQN and Graph Attention Networks (GAT) to address issues like random node connections and distance constraints for bandwidth-efficient use. By capturing intricate network relationships, GAT enhances node and link embedding, and Double-DQN dynamically optimizes resource allocation to improve overall performance and lower rejection rates for virtual network requests. As a reliable and scalable solution to the VNE problem, the proposed method outperforms traditional techniques in terms of results.