Graph Data Mining to Construct Sampled Internet Topology Maps

Abstract

Understanding the topological characteristics of the Internet is important for researchers and practitioners as the Internet grows with no central authority. This understanding is a necessity to better design, implement, protect and operate the underlying network technologies, protocols, and services. The need for accurate Internet topology map has increased recently with new services such as overlay networks and IP TV. Router-level Internet topology measurement studies have three main steps: topology collection, topology construction, and topology analysis. In topology construction, there are several main challenges: unresponsive router resolution, identification of underlying subnets and detection of IP aliases. These tasks become especially challenging when large-scale topologies of millions of nodes are studied. In this thesis, we present the topology construction processes of the Cheleby system, an Internet topology mapping system that provides insight into the Internet topology by taking daily snapshots of the underlying networks. The system utilizes efficient algorithms to process large-scale datasets collected from distributed vantage points and provides accurate topology graphs at link layer. Incorporating enhanced resolution algorithms, Cheleby provides comprehensive Internet backbone maps.