DRL-Driven Adaptive SFC Deployment with GCN in Distributed Cloud Networks

As cross-domain cloud networks become increasingly dynamic and complex, the deployment of service function chain (SFC) must adapt to varying quality of service (QoS) requirements and unpredictable traffic demands. In this paper, we propose an adaptive SFC deployment approach driven by deep reinforcement learning (DRL) integrated with graph convolutional network (GCN). By modeling the network as a markov decision process (MDP), the method captures dynamic state transitions and optimizes deployment strategies. Our approach incorporates advanced features such as a GCN-based encoder for network topology modeling, an LSTM for capturing sequential dependencies, and a customized reward function that balances latency minimization and resource utilization. Experimental results demonstrate the effectiveness of the proposed framework, achieving higher success rates, reduced latency, and improved system efficiency compared to traditional methods.