Context aware Elasticity Support for Virtualized Networks

Abstract

Mobile Operators are looking for new ways to cope with ever-increasing data traffic while improving the operational and capital efficiency of their networks. Cloud computing and network function virtualization (NFV) have emerged as key enablers to optimize resource utilization and at the same time reduce network operational expenditure (OPEX). In virtualized networks, network functions are delivered as software running on generic hardware allowing service providers to dynamically allocate resources based on traffic and service demands. This work presents resource utilization using real-life data of two different mobile networks and evaluate the impact virtualization would have on these networks.

Dynamic scaling of resources in NFV is a highly important challenge towards its implementation in real-life networks. In this work, a method to predict the required resources in the appropriate time to sustain true elasticity in NFV is presented. The capacity of different Virtualized Network Functions (VNFs) would increase/decrease in a way that the CPU utilization is maximized while the overall cost is minimized. Two strategies to predict the day-ahead CPU utilization are presented. The first strategy is an offline scheduling method that helps managing elasticity in virtualized networks by predicting normal days events. The second one is an online scheduling approach that predicts the day-ahead CPU utilization during sudden peaks due to some unusual circumstances. This work also presents new promising results that show the correlation between the control and data planes. Finally, a hybrid algorithm is proposed that uses both strategies to efficiently handle elasticity in virtualized networks. The obtained results are encouraging and are all based on real-life data of mobile operator networks.