Member： Malte Tashiro
Category: Exploring
- Background：The Internet is a continuously growing system of networks. The increasing structural complexity makes it difficult to understand all aspects of the system. In addition, the topology of the Internet can be represented in different ways, for example, as a map of physical cables, or more abstract as a mesh of connected networks.
- Purpose：Develop methods that infer dependencies from the Internet topology and increase knowledge about the Internet’s infrastructure by applying different levels of abstraction and performing a cross-layer analysis.
IXP Dependency Analysis
Internet exchange points (IXPs) play a vital role in the modern Internet. Envisioned as a means to connect physically close networks, they have grown into large hubs connecting networks from all over the world, either directly or via remote peering. It is therefore important to understand the real footprint of an IXP to quantify the extent to which problems (e.g., outages) at an IXP can impact the surrounding Internet topology. An IXP footprint computed only from its list of members as given by PeeringDB, or the IXP’s website, is usually depicting an incomplete view of the IXP as it misses downstream networks whose traffic may transit via an IXP although they are not directly peering there. In this paper we propose a robust approach that uncovers this dependency using traceroute data from two large measurement platforms. Our approach converts traceroutes to paths that include both autonomous systems (ASes) and IXPs and computes AS Hegemony to infer their inter-dependencies. This technique discovers thousands of dependent networks not directly connected to IXPs and emphasizes the role of IXPs in the Internet topology. We also look at the geolocation of members and dependents and find that only 3% of IXPs with dependents are entirely local: all members and dependents are in the same country as the IXP. Another 52% connect international members, but only have domestic dependents.
A more detailed description and additional results can be found in the related paper “Following the Data Trail: An Analysis of IXP Dependencies”.
Single-upstream Networks
Internet topology analysis based on BGP data only captures the active paths in the topology. This is because route collectors receive only preferred routes from their peers. As a consequence, it is hard to predict whether a failure of a network implies a failure of its dependents, or if the dependents have an alternative link available as a fail-over.
We assess Internet resilience by quantifying single-upstream ASes and studying their main characteristics. First, we analyze BGP data and show that single-upstream ASes account for half of the ASes on the Internet. Comparison between IPv4 and IPv6 shows that dual stack ASes are more likely to rely on a single upstream only in IPv6. Furthermore, if they have dependencies in both IPv4 and IPv6, they usually apply the same networking policy.
To validate these results, we perform active measurements to a random sample of 6366 ASes using traceroutes and BGP poisoning. We find that 72% of the ASes indeed rely on a single upstream, but also identify backup upstreams for 4%. Links to backup upstreams are normally not visible in BGP and only used if the primary upstream fails.
We characterize the measured ASes using different size metrics and AS type classifications and find that backup upstreams are usually only employed by smaller ASes, whereas reliance on a single upstream is seen for both small and large ASes.
Finally, we perform a case study of large eyeball networks covering 247 ASes from 163 countries for IPv4 and 141 ASes from 113 countries for IPv6. In total, 78% of analyzed countries in IPv4 and 81% in IPv6 have at least one large provider relying on a single upstream.
City Dependencies
The Internet is coupled to its underlying physical topology. Although this physical layer is transparent for routing purposes, the dependency between them should not be neglected.
We study how countries rely on cities in other countries for network connectivity by combining traceroute measurements with BGP and geolocation data. First, we map traceroute hops to cities and create a graph per target AS. Then, we compute Hegemony scores that represent how the AS relies on transit cities. Finally we aggregate ASes by country and weigh their combined Hegemony scores according to their announced IP space. We use this method to create four case studies.
- Maximum dependency per country: Shown in the figure above, we can distinguish between countries with high and low reliance on individual cities in other countries.
- Regional transit hubs: We identify countries (e.g., Singapore or Spain) that act as a transit hub for neighboring countries. We find that connection to different neighboring countries is handled by various cities.
- Submarine cable connectivity: The reliance on submarine cables is strongly reflected in city dependencies. Especially for connectivity to Africa, cities with landing points stand out.
- City dependency footpring: We can identify different characteristics for cities within a country by looking their dependencies. Most countries have one or more central city that is crucial for domestic connectivity, but only some countries have also cities with international reach.