Multi-Objective Collaborative Resource Allocation for Cloud-Edge Networks: A VNE Approach

The cloud-edge network (CEN) architecture has garnered significant attention due to its flexibility, reliability, and scalability in resource coordination and configuration. However, the generation of large-scale tasks has led to the urgent need for efficient resource allocation methods in CEN environments with limited computing resources. Virtual network embedding (VNE) technology enhances resource allocation flexibility by decoupling physical network resources and functions, allowing for adaptable integration of virtual networks (VNs) with underlying infrastructure. In this paper, we propose a deep reinforcement learning (DRL) based multi-domain VNE method, termed MD-VNE, for CEN resource allocation. Initially, the CEN is modeled as a multi-domain network with a series of associated resource constraints. Furthermore, we design an agent based on a multi-layer neural network to compute candidate CEN nodes and links. Finally, we validate the proposed method's advantages through extensive simulation experiments. The problem of efficient resource allocation in cloud-edge collaborative networks is effectively solved. Specifically, compared with the experimental baselines, the average improvements in the acceptance rate, long-term benefit and long-term benefit-to-cost ratio are $$, $$, and $$, respectively.