Thursday, September 19, 2002 Volen 101, 2:00-3.00 pm
In the past decade, the Internet has grown to reach every continent and nearly every nation. Yet, despite the network's astounding size and global reach, surprisingly little work has characterized the properties of this immense physical artifact. In this talk I will argue that we need to approach the study of the Internet's physical structure not as engineers but as scientists, looking at the Internet in a manner akin to a biologist studying a newly discovered organism.
In this spirit, I will describe our recent work studying the geographical location of Internet routers and links on the Earth's surface. Questions that motivate us are: What is the relationship between router placement and population density? How does distance affect the likelihood that two routers are directly connected? And: can we use the answers to these questions to improve generation methods for synthetic network topologies?
Our results suggest a superlinear relationship between router density and population density, across a wide diversity of geographic regions. Furthermore, the associated proportionality constant bears strong relationship to economic development. With respect to links, we find that over short distances, existing links show a marked distance-based preference, while at longer distances, link occurrence seems independent of distance. These results support one of the oldest (and least favored) synthetic network models -- the Waxman model. Finally, we combine router placement and link occurrence to suggest a new model for generating synthetic Internet-like topologies.
This is joint work with Anukool Lakhina, John Byers, and Ibrahim Matta.
Host: Liuba Shrira