ANALYSIS OF NETWORK RESOURCE SCALING MODELS IN 5G NETWORK

Abstract

This article analyzes the existing models of virtual network embedding and dynamic scaling of virtual network resources for Network Slices. These models make it possible to provide services with the required QoS while respecting the concept of efficient network resource usage. Dynamic virtual network embedding models allow virtual networks to be efficiently mapped on a substrate network and reconfigured on demand. But for more flexible dynamic scaling, Holt-Winters, Bi-LSTM and etc. models are additionally used, which are built according algorithms for predicting future resource utilization in order to reduce the initialization time of virtual network function instances serving certain Network Slices. Dynamic scaling models are compared and conclusions about the possibility of their use are given. As a conclusion we were made a summary about the possibility of using these models and the necessity of improvements for more flexible use in 5G networks.