Delay-Sensitive SFC Scheduling Optimization With DRL in Satellite-Terrestrial Networks.

Large-scale low Earth orbit (LEO) satellite communication systems play a pivotal role in shaping the future of 6G communication networks. This study delves into the intricacies of deploying delay-sensitive virtual network functions (VNFs) and scheduling service function chain (SFC) requests in satellite-terrestrial networks. We frame the SFC request scheduling (SFCRS) challenge as an integer linear programming (ILP) problem. We introduce an effective Deep Reinforcement Learning (DRL) approach to effectively tackle the SFCRS problem. The DRL-SFCRS algorithm primarily employs a sequence-to-sequence model and leverages Asynchronous Advantage Actor-Critic (A3C) for updating network parameters to attain the optimal policy. Simulation results underscore the algorithm's ability to adeptly balance the trade-off between end-to-end delay and resource utilization in the context of the SFCRS problem. As a result, the algorithm yields enhanced operator profits.