Towards Traffic Identification and Modeling for 5G Application Use-Cases

Abstract

To analyze next-generation mobile networks properly, there is a need to define key performance indicators (KPIs). Testing signaling only or just partial domains of the network have been replaced with end-to-end testing methodologies. With the appearing of machine-to-machine (M2M) applications, this question became even harder, since there is no direct user feedback. Quality of experience cannot be measured accurately in M2M applications, even if the network operates correctly and without failures. There are dozens of new—but theoretical—use-cases for 5G; however, these are not tested on a live network. The modeling methodology used throughout the paper follows the steps of observation, analysis, model creation, implementation, and verification. The first part of the paper examines the three application-types: enhanced mobile broadband (eMBB), critical Internet of Things (cIoT), and mass Internet of Things (mIoT). Afterwards, we introduce the main traffic characteristics based on current mobile networks’ traffic patterns and measurements. Considering the measurement results, we introduce a methodology and define traffic models for the simulation of different application-types. To validate these models, we compare the generated artificial traffic with real traffic patterns. In the second part of the paper, we examine what the main effects of these traffic patterns on a domestic 5G test-network are. Finally, we suggest some considerations on the possible main impacts regarding 5G network design.