Next: Experimental methodology
Up: Related work
Previous: Internet routing
Discovering and analyzing Internet structure has been the subject of
many studies. Much of the work has focused on studying topology purely
at the network level, without any regard to geography. Recently
several tools have been developed to map network nodes to their
corresponding geographic locations. A few Internet mapping projects
have used such tools to incorporate some notion of geographic location
in their maps.
The Mercator project [6] focuses on heuristics for Internet
Map Discovery. The basic approach is to use traceroute-like TTL
limited probe packets coupled with source routing to discover
routers1. A key component of Mercator is the set of heuristics used
to resolve aliases, i.e., multiple IP addresses corresponding to
(possibly different interfaces on) a single router. The basic idea is
to send a UDP packet to a non-existent port on a router and wait for
the ICMP port unreachable response that it elicits. In general,
the destination IP address of the UDP packet and the source IP address
of the ICMP response may not match, indicating that the two addresses
correspond to different interfaces on the same router. In our work we
use geographic information to identify points of sharing in the
network. We view this as complementary to network-level heuristics
such as the ones employed in Mercator.
The Internet Mapping Project [2] at Bell Labs also uses a
traceroute-based approach to map the Internet from a single
source. The map is colored according to the octets of the IP address,
so portions corresponding to the same ISP tend to be colored
similarly. The map, however, is not laid out according to
geography. Other efforts have produced topological maps that reflect
the geography of the Internet. Examples include the MapNet
[24] and Skitter [28] projects at CAIDA and the
commercial Matrix.Net service [25].
A number of tools have been developed for determining the geographic
location corresponding to an IP address. These tools use a variety of
approaches to map an IP address to location: inferring location from
Whois records [7] (e.g., NetGeo [11]),
extracting location information from traceroute data (e.g., GeoTrack
[13], VisualRoute [30]), determining the location
coordinates using delay measurements (e.g., GeoPing [13]),
etc. Our previous work on IP2Geo [13] focused on developing several
tools, including GeoTrack, to do IP-to-location mapping. In this work,
we use the GeoTrack tool to analyze geographic properties of Internet
routing.
Next: Experimental methodology
Up: Related work
Previous: Internet routing
Lakshminarayanan Subramanian
2002-04-14