Quantitative Analysis of the Internet Topology Model

Abstract

The article presents analysis of the Internet topology graph model. We use usual methods to analyse properties of the model but also propose some new characteristics. Our experimental data set comprises long-term traceroute records for several thousands of target hosts, repeatedly executed from a fixed source. We processed particular routes by merging them all together thus creating a single connected graph, which represents a sample of Internet-like network. We evaluate several network characteristics, notably node degree distribution, k-core decomposition, clustering coefficients, etc. Comparison shows good correspondence with other known results.Furthermore, we examine new metrics, particularly the count and length of linear subgraphs and the number of connections between vertices with maximum and minimum degree. These characteristics help to distinguish between network types. The latter one appears to be useful also in the context of finding a dominating vertex set problem.