Efficient Internet Topology Discovery Techniques

Abstract

Current macroscopic Internet topology discovery projects use large numbers of vantage points to conduct traceroute surveys of Internet paths. These projects send billions of unsolicited packets to millions of routers within the Internet. Due to the structure of the Internet, many of these packets are sent without gaining any new topology information. In this thesis, we implement and extensively test a largescale doubletree system designed to increase the efficiency of topology mapping projects and reduce the load that they place on the Internet. Also, for all of the effort that current projects put into gathering data, the methods used do not discover, with confidence, the entire set of paths. We propose, implement and critique a novel algorithm, economical MDA traceroute, which is designed to discover a comprehensive topology in a manner which is more efficient than the current state of the art. We show that, compared to current methods, well over 90% link coverage can be obtained while reducing the number of probes used by over 60%. We also evaluate alternate methods for making large scale topology discovery projects more efficient and comprehensive; such as using BGP routing data to guide probing.