<![CDATA[Virtual network embedding (VNE) is the effective mapping of virtual networks onto shared physical substrate networks while boosting resource utilization and ensuring quality of service (QoS). VNE is a real challenge in network virtualization, especially in the perspective of 6G-enabled datacenters, where the demand for ultra-low latency, heavy connectivity, and dynamic resource allocation is vital. The proposed solution enables the ability to infer indirect paths for the resources prediction task on the knowledge graph (KG) by making implicit meaningful relations among the entities that compose the resource network. The simulation results indicated the inference mechanism significantly improves efficiency and adaptability. This leads to overall performance gains in terms of runtime stability, resource utilization, and energy savings in dynamic 6G scenarios. The experimental results showed that the proposed solution provided a 24.9% reduction in energy consumption for small-sized virtual network requests (VNRs), while maintaining 24.8% and 23.9% for medium and large VNRs, respectively, while it significantly decreased the delay time compared to the resulted delay using the baseline models such as asynchronous advantage actor-critic (A3C) + graph convolutional network (GCN). The results also confirmed that the integration of the inference engine algorithm with the embedding process results in remarkable reduction in the execution time while preserving embedding accuracy.]]>
Copyrights © 2026
