1:
2: Internet topology analysis and modeling has attracted substantial attention
3: recently~\cite{FaFaFa99,WilRevisited,
4: TaGoJaShWi02,LiAlWiDo04,BuTo02,JaRoTo04,ZhoMo04}\footnote{We
5: intentionally avoid citing the statistical physics literature, where the number of
6: publications dedicated to the subject has exploded. For an introduction and
7: references see~\cite{DorMen-book03}.} because the Internet's topological properties and their evolution are
8: cornerstones of many practical and theoretical network research agendas.
9: Routing, performance of applications and protocols, robustness of the network under
10: attack, {\it etc.}, all depend on network topology. Since obtaining realistic topology
11: data is crucial for the above agendas, researchers have
12: focused on a variety of measurement techniques to capture the
13: Internet's topology.
14:
15:
16: Various sources of Internet topology data obtained using different
17: methodologies yield substantially different
18: topological views of the Internet. Unfortunately, many researchers either rely
19: only on one data source, sometimes outdated or incomplete, or mix disparate
20: data sources into one topology. To date, there has been little attempt to provide a
21: detailed analytical comparison of the most important properties of topologies
22: extracted from the different data sources.
23:
24: Our study fills this gap by analyzing and explaining topological properties
25: of Internet AS-level graphs extracted from the three commonly-used data sources:
26: (1)~traceroute measurements~\cite{skitter}; (2)~BGP~\cite{routeviews}; and
27: (3)~the WHOIS database~\cite{irr}.
28:
29: This work makes three key contributions to the field of topology research:
30: \begin{enumerate}
31: \item We calculate a range of topology metrics considered in the
32: networking literature for the three sources of data. We reveal the peculiarities of each data source and the
33: resulting interplay between artifacts of data collection and the key properties
34: of the {\em joint degree distributions} of the derived graphs.
35:
36: \item We analyze the interdependencies among an array of topological features
37: and observe that the {\em joint degree distributions} of the graphs define
38: other crucial topological characteristics.
39:
40:
41:
42: \item To promote and simplify further analysis and discussion, we
43: release~\cite{comp-anal} the following data and results to the community:
44: a)~the AS-graphs representing the topologies extracted from the raw data sources;
45: b)~the full set of data plots (many not included in the paper) calculated for
46: all graphs;
47: c)~the data files associated with the plots, useful for researchers
48: looking for other summary statistics or for direct comparisons with
49: empirical data;
50: and d)~the scripts and programs we developed for our calculations.
51: \end{enumerate}
52:
53:
54: We organize this paper as follows. Section~\ref{sec:data} describes our data
55: sources and how we constructed AS-level graphs from these data.
56: In Section~\ref{sec:characteristics} we present the set of topological
57: characteristics calculated from our graphs and explain what they measure and
58: why they are important. We conclude in Section~\ref{sec:conclusion} with a
59: summary of our findings.
60: