Instantaneous Bandwidth Approximation in 5G Networks

Abstract

The estimation of available bandwidth in real-time is a challenging problem which has resulted in various proposed solutions for various network architectures. Because network conditions can change quickly, it can be difficult to estimate the available bandwidth without measuring it. Instead of directly measuring data speeds, which can use up the network’s capacity and be slow, another approach is to use counters that exist in 5G base stations. These counters track network activity and offer a simpler way to estimate bandwidth without putting extra load on the system. This thesis aims to develop a model for estimating bandwidth in base stations within 5G networks. It also explores the selection of appropriate counters for this purpose, examines potential correlations between them, and investigates how various traffic scenarios impact the accuracy of bandwidth estimation. A quantitative methodology is used to this end. There has been previous research done on the topic of bandwidth estimation and prediction, however not as much has been done using counters as input. The results show that certain counters related to data volumes and resource block symbols appear to be particularly well suited for estimating bandwidth. The correlations themselves suffice to be simple linear relationships and are for the most cases accurate. The models estimate differently well for different traffic scenarios. The scenarios with two devices transmitting data overall have higher error rates than comparing to the one device scenarios.