On the Semantics of Internet topologies

Author(s)
Mihail, Milena
Gkantsidis, Christos
Saberi, Amin
Editor(s)
Associated Organization(s)
Organizational Unit
Supplementary to:
Abstract
To date, the most successful approach to modeling the AS-level topology is the degree-driven approach of Inet. Inet predicts the degrees of the topology by extrapolation from available data, then constructs a topology meeting the degree sequence using a preferential connectivity heuristic. We focus on two important areas left open by prior work. First, we explore the theoretical foundations of degree-based graph generation. We identify the relevant results from graph theory, and exploit these to improve fundamental understanding and produce richer models. Second, essentially all prior AS-level models have the characteristic that they contain extremely limited {\em semantics}. The graphs produced are undirected and unlabeled, hence they simply reflect connectivity without any notion of additional semantic information. We address the issue of adding semantics to network topologies, in the areas of peering relationships and clustering into higher-level groupings based on geographic and/or business relationships. Our techniques for adding semantics suggest new methods for evaluating the quality of generated topologies.
Sponsor
Date
2002
Extent
321058 bytes
Resource Type
Text
Resource Subtype
Technical Report
Rights Statement
Rights URI