Efficient Autoscaling of VNF(s) in NFV environment

Abstract

Autoscaling is one of the principal objectives for the intent based future networks. Scaling the cloud resources in different situations was dependant on a cloud expert administrator. This dependancy led to the demand of automating the process of scaling resources. Specifically, one of the key players that drove and came up with the idea of autoscaling network resources was SDN, Invention of SDN led to the possibility of autoscaling in cloud environments because of its capability to separate the control plane from the data plane. Automating the process of scaling network resources was not possible without the need of an administrator but as the networks evolved with the invention of SDN, it led to provide connectivity between the virtualized network resources. To use the network resources efficiently is the focal interests of future networks. The proposed system comprises of an intent based autoscaling application, monitoring microservice and a few modifications proposed to the elementary management services. The application takes the network resource's key factors into account such as cores in execution-times, CPS, RAM, Hard disk, hyper-threads etc, and keeps weight factor into consideration for virtualized/physical core, while deciding the resource scalability after the intervals. All the factors mentioned above, ensure to estimate the assignation of resources up to optimal. This mechanism provides us a platform with efficient computation capabilities. While the necessary monitoring keeps track of the information from all over the virtual network resources assigned to a specific tenant. The proposed monitoring service being at the management layer provides us a way to fetch realtime data from the resources in a less latent approach. For the overall system, M-CORD (a Framework developed by ONF) was used to deploy network functions inside compute machines and integrated the proposed system with it to evaluate the aforementioned key factor such as efficiency, in terms of CPU, RAM usages for specific intervals.