Background

As the Internet has grown, so has the challenge of accurate measurement and modeling of its topology. Commonly used but coarse methods of measuring topology, e.g., BGP tables, suffer from several limitations. To pursue more accurate empirically-based topology modeling, CAIDA began its Macroscopic Topology Project in 1998. The project focus is actively measuring topology and round-trip time (RTT, the time interval between the moment a probe is sent and the moment a response is received) across a wide cross-section of the commodity Internet.

For further information, see the paper "Topology discovery by active probing", the tool skitter, and CAIDA's analysis page.

Description

CAIDA's topology monitors send out traceroute-like packets to as many as several hundred thousand destinations everyday while measuring round-trip time and recording the forward path taken by the packets. Our visualization shows a single cycle of measurements made by the monitor in Herndon, VA on Feb 2, 2002. From this data, we created a graph showing the topology of the Internet covered by the probes. We then overlaid RTT measurements on the links. Since links, properly speaking, do not have RTTs, and since the monitor collects RTTs to destinations only (and not also to the routers lying along the path), we colored a link by the median of the RTTs seen for all destinations that were reached through the link. Thus the color of a link summarizes the performance characteristics of all destinations that fan out from the link. Clusters of one color indicate clusters of destinations with similar latency values from this source monitor. This coloring also allows one to see the intermediate nodes at which performance begins to diverge.

Geographical location has some effect on the RTTs one can expect from a destination. As the following plot (based on a different set of data collected from San Diego, CA) shows, there is a rough correlation between physical distance and RTT. The RTTs exhibit a trimodal distribution in which the peaks correspond to the east and west coasts of North America and to Europe/Asia. In the Walrus visualizations further below, the cyan links, which indicate the lowest RTTs, are most likely to destinations on the east coast where the source monitor is located. The green links are most likely to destinations on the west coast, and the yellow links to destinations in the rest of the world. The red links, indicating RTTs greater than 300ms, are likely showing poor or misbehaving connectivity.

Visualizations

Click on an image for the larger 2048x2048 (464K) version, or click on the text below each image for the larger version without an accompanying key.

Each image uses the following key:

[Click for a version without the key.]

[Click for a version without the key.]

Credits

Visualization: Young Hyun, youngh@caida.org
Data Analysis: Bradley Huffaker, brad@caida.org
Walrus is a CAIDA tool developed with support of DARPA NGI N66001-98-2-8922, DARPA NMS N66001-01-1-8909, NSF ANI-9996248, NSF N66001-01-1-8909, and the support of CAIDA members.
CAIDA is based at the University of California's San Diego Supercomputer Center.