Evaluation of a Large-Scale Topology Discovery Algorithm

Benoit Donnet
Université Pierre & Marie Curie
Laboratoire LiP6/CNRS, UMR 7606, France
and
Cooperative Association for Internet Data Analysis - CAIDA
San Diego Supercomputer Center,
University of California, San Diego

Bradley Huffaker and kc claffy
Cooperative Association for Internet Data Analysis - CAIDA
San Diego Supercomputer Center,
University of California, San Diego

Timur Friedman
Université Pierre & Marie Curie
Laboratoire LiP6/CNRS, UMR 7606, France

Abstract. In the past few years, the network measurement community has been interested in the problem of internet topology discovery using a large number (hundreds or thousands) of measurement monitors. The standard way to obtain information about the internet topology is to use the traceroute tool from a small number of monitors. Recent papers have made the case that increasing the number of monitors will give a more accurate view of the topology. However, scaling up the number of monitors is not a trivial process. Duplication of effort close to the mon itors wastes time by reexploring well-known parts of the network, and close to destinations might appear to be a distributed denial-of-service (DDoS) attack as the probes converge from a set of sources towards a given destination. In prior work, authors of this paper proposed Double tree, an algorithm for cooperative topology discovery, that reduces the load on the network, i.e., router IP interfaces and end-hosts, while dis covering almost as many nodes and links as standard approaches based on traceroute. This paper presents our open-source and freely download able implementation of Doubletree in a tool we call traceroute@home. We evaluate the performance of our implementation on the PlanetLab testbed and discuss a large-scale monitoring infrastructure that could benefit of Doubletree.