Network Traffic Prediction for a Software Defined Network based Virtualized Network Functions Platform

Abstract

Network Functions Virtualization (NFV) separates the software implementation of network functions from the physical infrastructure and allow them to be offered as Virtualized Network Functions (VNFs). When offering VNFs as services, the Cloud Service Providers (CSPs) are responsible for handling demand-oriented resource provisioning, network access, security, etc. CSPs monitor the resource utilization and network traffic to maintain a global view of the network in terms of its resources provisioning and network congestion. With the use of Software Defined Network (SDN) architecture in the infrastructure, CSPs can easily perform monitoring processes, from the central point: the SDN controller. Based on the data gathered from the monitoring process, CSPs performs traffic predictions for the future, so that they can ensure the allocation of required resources, specially by scale in/ out to avoid any over-provisioning or under-provisioning. In this paper, we have explored Machine Learning approaches that can be used to predict the traffic of a cloud platform that uses an SDN architecture and offers VNFs as services. We have focused on Facebook (FB) prophet approach and Random Forest (RF) regression models. Our results show that the RF model showed an accuracy of 86.7%, and FB prophet can predict the traffic for different time periods, by identifying trends for the year, month, and week.