• data visualization
• inter-domain
• internet
• monitoring
• network measurements
• performance
• routing

The Internet is a critical infrastructure in our society. The collection of data about its functioning and the enrichment, analysis, and visualization of such data are activities of paramount importance, from both a research and an operational point of view. The routing is what allows data to flow between devices connected to the Internet. At any time, the protocols that make the Internet work can establish different network paths to follow in order to deliver such data. Revealing routing behaviors allows operators to design, maintain, and diagnose the Internet and enables researchers to better understand it in order to improve it.

In this thesis, we present several approaches for the analysis and visual representation of Internet routing data.

We initially focus on the geolocation of routers and servers that make the Internet work. We describe and evaluate an active geolocation service, called RIPE IPmap. This system is able to estimate the location of a connected device by performing latency measurements towards it. We then study several factors influencing active geolocation processes, and provide an estimation of the maximum theoretical accuracy achievable worldwide by a generic active IP geolocation system, establishing in this way a baseline for future research on the topic.

Further, we tackle the problem of supporting network operators and researchers performing routing data analysis by designing and testing new visualization metaphors that can automatically represent such data. In particular, we introduce Radian, a tool implementing a new visual metaphor for the representation of evolving network topologies. We develop new algorithms for drawing dynamic clustered graphs evolving over time, able to support important cognitive aspects during network operations. We then focus on the inter-domain routing case and present Upstream Visibility, a visualization supporting the monitoring of network reachability. While doing so, we define and experiment with several heuristics, contributing to the generic problem of automatically generating readable stacked area chart diagrams.

Finally, we apply the methodologies devised to analyze some Internet behaviors at scale. First, we study phenomena of periodic topology changes in network measurement data and identify the causes behind some of them. Second, we analyze the routing situation in the Middle East, characterized by high latencies caused by neighboring countries not connecting with each other. Third, we perform the first large-scale evaluation of the impact of the COVID-19 pandemic on Internet latency, showing how the increased amount of online activities affected Internet performance.