Application of Machine Learning and Deep Learning Algorithms in Predicting Virtual Network Functions for Network Function Virtualization

Network functions virtualization (NFV) is a methodology that replaces hardware middleboxes with more flexible software alternatives called virtualized network functions (VNFs). These VNFs need to be created and their resource allocation needs to be changed as per traffic demands to dynamically adapt to changes. However, determining the appropriate number of resources to be allotted for each VNF instance is challenging, as fixed resource consumption is often assumed in existing optimization methods. This can lead to either resource wastage or poor service quality. A novel machine learning algorithm trained on actual VNF data, which includes performance measures and resource needs is proposed to address this issue. These trained models can accurately predict the number of resources needed for each VNF to handle a specific traffic load. These machine learning models are incorporated into an algorithm for simultaneous VNF placement and scaling, and their effect on the final VNF placements is evaluated.