The contents of this legacy page are no longer maintained nor supported, and are made available only for historical purposes.

Bibliography Details

Y. Hyun, A. Broido, and k. claffy, "On Third-party Addresses in Traceroute Paths", in Passive and Active Measurement Conference (PAM), Apr 2003.

On Third-party Addresses in Traceroute Paths
Authors: Y. Hyun
A. Broido
k. claffy
Published: Passive and Active Measurement Conference (PAM), 2003
URL: https://catalog.caida.org/paper/2003_3rdparty/3rdparty.pdf
https://catalog.caida.org/paper/2003_3rdparty/3rdparty.ps.gz
https://www.caida.org/catalog/media/2003_pam0304/pam2003-3rdparty.pdf
Entry Date: 2003-08-13
Abstract: Traceroute IP paths are often used in studies of Internet topology and routing. Though producing one of the best available router-level maps of forward paths, traceroute is susceptible to several types of inaccuracies. However, no one to date has quantified the magnitude of these inaccuracies in real-world traceroute paths. We make an initial attempt by reporting on the observed frequency of one type of inaccuracy--third-party addresses. A third-party address is the address of a router interface that does not lie in the actual path taken by packets. Based on an examination of thousands of traceroute paths from six locations worldwide and the application of several metrics, we find that the situations in which third-party addresses can occur to be relatively uncommon. They mostly occur within a few hops of the destination (that is, at the destination edge of the network), with multihoming being their most likely cause. Our conclusion is that third-party addresses cannot be a significant source of AS map distortions.
Datasets: Several hundred thousand traceroute paths from six skitter monitors (3 in US, 2 in Europe, 1 in Japan) taken Jan 10-13, 2003.
Results:
  • 3rd-party addresses that cause incorrect AS paths relatively uncommon
  • 3rd-party addresses tend to occur near the destination
  • multihoming can lead to 3rd-party addresses
  • impact on AS-level analysis probably small