DRL-Based Dynamic SFC Orchestration Algorithm for LEO Satellite Networks

Low-Earth-Orbit (LEO) satellite networks face unique challenges in service function chain (SFC) orchestration due to their dynamic topology and resource constraints, making traditional terrestrial network solutions inadequate. This study addresses the challenge of maximizing service provider benefits through efficient SFC deployment and readjustment in LEO satellite networks. We propose PPOSFC, a novel dynamic SFC orchestration algorithm based on Proximal Policy Optimization (PPO), which incorporates future topology information into the decision-making process. The algorithm models the orchestration problem as a Markov decision process and employs a dual-objective optimization approach considering both deployment success rate and readjustment costs. Simulation results demonstrate that PPOSFC achieves a 13.07% increase in cumulative profit and improves deployment success rates by 6.78% compared to existing algorithms. The algorithm exhibits superior performance in both high and low user service request intensity, effectively balancing service quality and operational efficiency. Furthermore, our analysis reveals that incorporating predicted topology information significantly enhances orchestration performance.