SFC Embedding Meets Machine Learning: Deep Reinforcement Learning Approaches

Service function chain (SFC) has been recognized as one of the most important technologies that can satisfy dynamic service demands in the edge clouds. However, how to efficiently embed SFCs in the dynamic edge-cloud scenarios remains as a challenging problem. Considering different network topologies, we devise two deep reinforcement learning (DRL)-based methods for two network sizes: a deep deterministic policy gradient (DDPG) based method for the small-scale networks and an asynchronous advantage actor-critic (A3C) based approach for the large-scale networks. Simulation results demonstrate that our proposals can efficiently deal with the SFC-DMP in edge clouds and outperform the state-of-the-art methods in terms of the delay.

Automated summary

The service function chain (SFC) is a crucial technology for dynamic service demands in edge clouds, and two deep reinforcement learning-based methods have been proposed to efficiently embed SFCs in different network sizes, outperforming existing methods in terms of delay according to simulation results.