Intercontinental Internet communications lie on top of a complex network of submarine cables forming its global communication backbone. As part of ongoing work focused on the criticality of the submarine cable network, fellow researchers from Northwestern University and the University of Wisconsin-Madison and I mapped traceroute measurements to submarine cables as a first step towards understanding its potential vulnerabilities.
We initially set out to follow an approach first introduced in a position paper, building on the assumption that given a traceroute, one could identify the hop or link with the largest latency and if the associated routers were found near submarine landings, that this link could be mapped to one or a handful of cables. Unfortunately, a preliminary analysis of traceroute datasets changed this assumption.
While we found traceroutes matching these expectations, we also found many others in which the routers associated with a submarine-traversing link were far inland from the closest landing points, some as far as 700kms. Further analysis revealed many of these long-haul links covering distances larger than 10,000kms and connecting every economy worldwide.
Motivated by these early observations, we conducted a longitudinal study of intercontinental long-haul links and their preferred destinations. In a paper to appear at SIGMETRICS 2024, we report on the first-ever study of this long-haul infrastructure using a large corpus of traceroute data collected at the network’s edge and spanning over seven years.
Our study revealed a vast network with links spanning thousands of kilometres and central nodes connecting as many as 45 economies.
While long-haul links can be thought of as the network-layer manifestation of critical transoceanic cables, we find that long-haul links are substantially longer than any intercontinental submarine cable segment, with a median Round-Trip Time (RTT) of 130ms – nearly 84% larger than the median RTT of submarine cable segments (70.76ms).
The termination points of these links are located quite far from the nearest, on-route, landing points. Specifically, 64% of them are at a distance of 500kms from the nearest landing point, and in 10% of the cases, that distance exceeds 3,513kms. Chicago, a popular ‘destination’ is more than 1,000kms away from the nearest landing point and yet connects over 60 economies within one hop.
Our study contributes to the community effort to create consistent maps across layers of the Internet, from AS-level to logical and physical connectivity, critical to a range of important analyses from performance and robustness to security.
This new perspective opens a wide range of promising directions for future research, from alternative views of the long-haul infrastructure to exploring that infrastructure’s key properties and temporal stability.
To learn more about our study read our paper ‘A hop away from everywhere: A view of the intercontinental long-haul infrastructure,’ which we will present at the 2024 ACM SIGMETRICS Conference.
Esteban Carisimo is a Postdoctoral Researcher at AquaLab, a group at the Department of Computer Science of Northwestern University, whose research takes an interdisciplinary approach to Internet topology, including IXPs, CDNs, transit diversity, and network congestion.
Contributors: Fabián E. Bustamante (Northwestern University), Caleb J. Wang (Northwestern University), Mia Weaver (U. of Wisconsin-Madison) and Paul Barford (U. of Wisconsin-Madison).
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.
Fascinating research! I’d love to hear more about your findings on the submarine cable network’s vulnerabilities and how they impact global internet communications.