An Improved Adaptive Service Function Chain Mapping Method Based on Deep Reinforcement Learning

With the vigorous development of network functions virtualization (NFV), service function chain (SFC) resource management, which aims to provide users with diversified customized services functions, has gradually become a research hotspot. Usually, desired by user is randomness and timeliness, formed request (SFCR) dynamic real-time, requires that SFC mapping can be adaptive satisfy dynamically changing requests. In this regard, paper proposes an improved method based on deep reinforcement learning (ISM-DRL). Firstly, model proposed abstract process decompose problem into SFCR VNF reorchestration problem. Secondly, we use deterministic policy gradient (DDPG), framework, jointly optimize effective cost rate approximate optimal strategy for current network. Then, design four orchestration strategies rate, etc., enhance matching degree ISM-DRL different networks. Finally, results show in realize processing under request. Under experimental conditions, performs better than DDDPG DQN methods terms average utilisation rate.