CAIDA : data : active : as-relationships

Background

The Internet is composed of thousands of ISPs that operate individual parts of the Internet infrastructure. ISPs engage in both formal and informal relationships to collectively and ubiquitously route traffic in the Internet. These relationships are usually realized in the form of business agreements that translate into engineering constraints on traffic flows within and across individual networks participating in the global Internet routing system.

Accurate data on the structure of actual relationships among ASes is required for many research efforts concerned with performance, robustness, and evolution of the global Internet. Examples of both research and operational tasks that cannot neglect AS relationships include:

realistic simulations trying to model path inflation effects caused by routing policies;
understanding how packets are routed in the Internet and how to optimize Internet paths by analyzing existing deficiencies;
development of more scalable interdomain routing protocols and architectures, like HLP, that take into account the structure of AS relationships to optimize their performance;
evaluating how AS relationships affect the evolution of the Internet infrastructure and constructing economy-based models of the global Internet growth;
analysis of the spectrum of BGP configuration scenarios in order to develop more expressive routing protocols and configuration languages;
inference of AS paths in the Internet;
modeling the structure of routing tables and developing synthetic routing tables needed for simulations of routing table lookup algorithms;
development of better topology generators that account for the topological idiosyncrasies associated with AS relationships;
selection of data centers for server replicas by measuring the origin of traffic to existing servers and evaluating connectivity and AS relationships of candidate data centers; and
selection of peers or upstream providers based on connectivity and AS relationships of candidate ISPs.

Inferred AS Relationships Dataset

Download the AS links annotated with AS relationships Dataset

Please cite this dataset as follows:

The CAIDA AS Relationships Dataset, <date range used>. http://www.caida.org/data/active/as-relationships/

For more information, email topology-info@caida.org.

Data is available from 2004 to present, with one file created per week in 2006 and one per month in prior years. Each file contains a full AS graph derived from RouteViews BGP table snapshots taken at 8-hour intervals over a 5-day period. The AS relationships available are customer-provider (and provider-customer in the opposite direction), peer-to-peer, and sibling-to-sibling. See the comments at the beginning of each file for details of the file format.

The general procedure for creating a file is as follows:

Extract all AS links from RouteViews snapshots.
Infer customer-provider relationships, and annotate AS links.
Infer peer-to-peer relationships, and annotate AS links, possibly overriding customer-provider relationships inferred in step 2.
Heuristically fix suspicious looking inferred relationships (e.g., a low-degree AS acting as provider to a high-degree AS).
Infer sibling ASes (that is, ASes belonging to the same organization) from WHOIS, and annotate AS links, possibly overriding previous relationship annotations.

For details of the algorithms used to infer AS relationships, see the following papers:

Related Work

AS relationships are used to determine the ranking of ASes based on the size of customer cones. The rankings are available through the AS rank CGI.
In the paper Revealing the Autonomous System Taxonomy: The Machine Learning Approach, AS relationships play an important role in classification of ASes into the following broad categories: large ISPs, small ISPs, customer networks, university networks, Internet exchange points, and network information centers. An empirical AS taxonomy based on these categories is available for download at the Autonomous System Taxonomy Repository page.