Machine learning based scalable and adaptive network function virtualization

Abstract

Due to the continuous development of SDN and NFV technology in recent years, it is important to improve the network performance to the users. But the traditional technology is not mature and has many shortcomings. In order to apply these two technologies (Adaptive and Autoscaling) in computer networks, we use SDN not only to separate the forwarding plane and control plane, but has the nature of the programmability also. Based on the actual business requirements for automatic deployment, NFV technology has the resources of virtualization and the characteristics of flexibility and fault isolation. Two kinds of technology are different, but they can work cooperatively very well. In this paper, we apply the algorithms of machine learning, combine the SDN and NFV technology, and build NFV dynamic control system architecture on the CloudStack cloud platform to provide users with customized service. Furthermore, in the fourth part, we added the feasibility of architecture to the home network and mobile core network.