Energy Efficient VNF Placement Algorithm Using Reinforcement Learning in NFV-Enabled Network

The application of network function virtualization (NFV) allows current networks to flexibly manage orchestrated network functions and provide various network services. The current virtual network function placement process suffers from excessive energy consumption and high cost, and reducing energy consumption is one of the important challenges in the NFV resource allocation process. In this article, we explore the issue of placing VNFs within networks that have been enabled for NFV. We propose an algorithm for energy-efficient deployment of virtual network functions based on improved reinforcement learning. NeMC combines neighborhood node information integration with Monte Carlo Tree Search (MCTS) method while evaluating the relevance of placed nodes. We set the underlying node resource occupation criteria for the placement policy and dynamically shut down the unused nodes to significantly reduce energy consumption; link selection utilizes the shortest path search technique to achieve end-to-end path search. Evaluation results show that the solution performs well in handling many online NFV requests, with reduced energy consumption compared to the comparison algorithm. The benefit-cost ratio is also improved.