DRL-Based Service Function Chains Embedding Through Network Function Virtualization in STINs

Space-terrestrial integrated networks (STINs) are gaining increasing attention for their outstanding benefits in providing seamless connectivity, enhancing network resilience, increasing capacity, and expanding coverage. The combination of network function virtualization (NFV) and deep reinforcement learning (DRL) is a promising candidate to boost the capability of STINs to deliver high-quality services. In this paper, a STIN architecture based on artificial intelligence (AI) algorithms and the process of embedding Service Function Chains (SFCs) into highly dynamic STINs are studied. Then, we propose an enhanced DRL-based SFCs embedding algorithm that initiates with a feature fusion phase of links, followed by aggregating the feature vectors of all nodes and their neighboring nodes to form context information. This context information is then used as part of the input state for the DRL-based SFCs embedding algorithm. Simulation results show that the proposed algorithm can effectively reduce the latency of SFCs embedding when compared with other existing algorithms, thus showing the superiority of our proposed algorithm.