🎉 Our paper was accepted by ACM CoNEXT 2024!

Table of Contents

Author

Yibo Ma (Me), Tong Li, Yuwei Du, Schahram Dustdar, Zhaocheng Wang, Yong Li (Supervisor, liyong07@tsinghua.edu.cn)

Abstract

Although 5G networks offer larger capacity due to more antennas and larger bandwidths, their increased energy consumption is concerning. This paper investigates energy consumption issues from widespread 5G deployment using city-scale real-world mobile network data. Our dataset includes traffic volume, energy consumption, and base station attributes spanning May 2022, July 2023, and April 2024, covering over 10,000 4G and 5,000 5G base stations. A portion of the dataset is published on Google Drive. We develop high-accuracy models to profile 4G and 5G base station energy consumption, revealing 5G inefficiencies under low traffic loads. We identify energy efficiency traps where network energy efficiency drops post-5G deployment. Our study estimates the recovery time from energy efficiency traps, emphasizes that different functional regions have varying recovery levels, and analyzes the reasons for these differences. We provide recommendations for deploying and commercializing 5G networks to promote sustainable 5G connectivity.

Dataset

We gathered a dataset of measurement reports, including traffic volume, energy consumption, and base station attributes for city-scale mobile networks. The dataset spans three periods: May 2022, July 2023, and April 2024. It includes data from over 10,000 4G base stations and 5,000 5G base stations, capturing the complexity of real-world network operations. To support the research community, we have published a portion of the dataset, including complete record traces of all base stations over one weekday and one weekend, on Google Drive.

Acknowledgements

This research has been supported in part by the National Natural Science Foundation of China under Grant U22B2057 and Grant U20B2060; in part by the National Key Research and Development Program of China under Grant 2020YFA0711403.

License

Released under the MIT license.