Paths in the cloud: Geolocation mapping of Amazon's cross-border connectivity

Jinyu Zhu, Yu Zhang, Yunan Wang, Hongli Zhang, Binxing Fang

Abstract: In the evolving landscape of the global digital economy, the growing demand for cloud services is driving significant expansion of cloud infrastructure worldwide scale. To safeguard cloud data security and comply with regulatory requirements, this paper presents a geographic perspective to study the internal topology of Amazon’s Cloud network. We introduce a cloud-centric prefix measurement methodology designed to capture the internal network topology of the cloud provider with greater precision. This method facilitates the identification of national border IP addresses and cross-border links using geolocation data, providing a comprehensive understanding of the cloud infrastructure’s international connectivity. Compared to traditional global internet measurement techniques, our method discovers 6.87% more new border IP addresses, covers 11.11% additional countries and extracts 31.85% more inter-country connections at a probe cost of 31.50% and a time cost of 26.23%. For the first time, we present the internal network structure of Amazon Cloud and analyze cross-border connections. The United States has the highest number of border IP addresses, connecting with all other countries, followed by Ireland. In China, we only found border IP addresses and cross-border links within Hong Kong.