Virtual Network Function Deployment Algorithm Based on Q-learning in MEC

As a new cloud service mode, mobile edge computing distributes the traditional centralized deployment and management of cloud resources to the wireless access network, so that mobile services can be processed nearby, so as to obtain good service experience and reduce the network load of the backtrip network. In order to effectively improve the end-to-end service delay of the flow in multi-clusters coexisting mobile edge computing (MEC) network, an improved virtual network function deployment method based on Q-learning algorithm was proposed. Based on the planning model, a spatial-temporal optimization model of service chain deployment was established by markov decision process, and an improved BPQ-learning algorithm was designed to solve the model. The method considered both the virtual mapping of MEC service chain in space dimension and the life cycle management of VNF in time dimension. The optimization of VNF deployment is realized. The possible network congestion is avoided by mapping service nodes in advance. Experimental results show that the proposed strategy can provide end-to-end services with lower latency and better experience for delay-sensitive mobile services under different service request volume, service node scale, cluster number and logical connection relationship among virtual network functions. © 2022, Taiwan Ubiquitous Information CO LTD. All rights reserved.