A power-efficient and performance-aware online virtual network function placement in SDN/NFV-enabled networks

Recent development of software-defined networks (SDN) and network function virtualization (NFV), makes it feasible to replace dedicated hardware middleboxes with software virtualization to run network functions on general-purpose servers. The main challenge in the NFV is virtual network function placement (VNF-P) which is an NP-hard problem, and thus, heuristics and metaheuristics can be used to solve it. In this paper, a hybrid heuristic-metaheuristic learning model is introduced for online VNF-P. In this model, a multi-criteria heuristic (named MCH) is utilized for online VNF placement and routing, while a metaheuristic based on genetic algorithm (GA) is applied in an offline procedure to learn the hyperparameters of the online MCH model, aims to minimize the total power consumption in the NFV infrastructure. The optimization procedure using GA is performed only once before the main operation of the NFV. Simulation results demonstrate the effectiveness of the MCH-GA learning model against the existing methods.