On the Incompleteness of the AS-level graph: a Novel Methodology for BGP Route Collector Placement

In the last decade many studies have used the Internet AS-level topology to perform several analyses, from the discovery of its graph properties to its impact on the effectiveness of worm-containment strategies. Yet the BGP data typically used to reveal the topologies are far from being complete. Our contribution is three-fold. Firstly, we analyze BGP data currently gathered by RouteViews, RIPE RIS and PCH route collectors, and investigate the reasons for its incompleteness. We found that large areas of the Internet are not properly captured due to the geographic placement of the feeders and to BGP filters, such as BGP export policies and BGP decision processes. Secondly, given the incompleteness of data available, we design a novel algorithm for selecting the optimal number of ASes that should join a route collector project to obtain a view of the Internet AS level topology that is closer to reality. We applied this algorithm to the global topology and to five regional AS-level topologies, highlighting that the particular characteristics of the Internet at a regional level cannot be ignored during this process. Thirdly, we provide a characterization of the ASes that we found to be part of at least an optimal solution set, and highlight the characteristics of a typical AS that should join a route collector project. By analyzing these ASes we found that the current route collector infrastructure is rarely connected to them, highlighting that much more effort should be made in devising an ideal route collector infrastructure that ideally would be able to capture a complete view of the Internet.