An A3C Learning Approach for Adaptive Service Function Chain Placement in Softwarized 5G Networks

Abstract

Network functions virtualization transforms traditional network functions into software, enabling them to run as virtual network function (VNF) instances on cloud infrastructure. In softwarized 5G networks, communication services are provided through service function chains (SFCs), which sequentially link multiple VNFs according to specific requirements. This approach enhances network management and orchestration, offering greater flexibility and scalability. However, improving resource consumption and quality of service while adhering to physical network constraints remains a significant challenge. This study introduces an A3C-GLA framework for adaptive service function chain placement (that leverages the Asynchronous Advantage Actor Critic (A3C) algorithm, Graph Attention Networks (GATs), and Sequence-to-Sequence Long Short-Term Memory with Attention mechanism (Seq2SeqLSTM-A). Extensive simulations demonstrate that the proposed framework significantly outperforms existing benchmark schemes in terms of long-term average revenue and acceptance ratio, offering a more efficient and effective solution for SFC placement in softwarized 5G networks.