Multi-criteria reactive approach for joint dynamic VNF load balancing and service auto-scaling in NFV

The evolution of 5G key-enabler technologies, such as Software-defined Networking (SDN) and Network Function Virtualization (NFV), has brought network operators" attention to procure an efficient service delivery mechanism. Therefore, it is vital to leverage the service management and orchestration functionalities that operate in harmony. This can curb the undesirable negative impact of inconsistency on the quality of service. In this paper, we investigate the joint load balancing and auto-scaling of the elastic services that are being provisioned in the computing infrastructure at the edge or cloud. We address the necessity and challenges of designing the load balancing algorithm and scale decision making policy, aware of one another, through several practical scenarios on multimedia service delivery in NFV. It is demonstrated that changing the VNF load balancing method may deteriorate the scaling quality of decision-making. Consequently, we propose a novel multi-criteria reactive approach using a mutual score to dynamically adapt the VNF load balancing algorithm with the auto-scaling engine. We implement the proposed joint method on the management and orchestration layer of our previously developed NFV/SDN testbed, called XeniumNFV, to evaluate the effectiveness of our approach by conducting extensive experiments on the elastic web service.