Abstract:

The paper concerns resource management in the Software-Defined Wide Area Networks (SD-WANs). In the context of SD-WAN, SDN principles are extended to manage and optimize traffic across geographically dispersed networks, crucial for large enterprises and service providers. The paper highlights the Virtual Network Embedding (VNE) problem, which involves mapping virtual network nodes and links to physical network resources, ensuring efficient resource allocation, and meeting Quality of Service (QoS) requirements. The proposed solution framework, the Automatic Resource Managing System (ARMS), leverages machine learning for problem classification and algorithm selection, enhancing resource management automation in SDNs. The paper also explores the application of Vertical Federated Learning (VFL) in multidomain SD-WANs, allowing collaborative yet privacy-preserving resource management across different network domains. Initial experimental results demonstrate the feasibility of the proposed approach, showing promising accuracy, precision, and recall metrics for algorithm selection in VNE problems.