Check out the new USENIX Web site. next up previous
Next: Future Work Up: Mapping and Visualizing the Previous: Watching Networks Under Duress

Related Work

There are a number of Internet data collection and mapping projects underway. Some have been running for a number of years, such as John Quarterman's Matrix Information and Directory Services [18], which includes the ``Internet weather report.'' Martin Dodge has collected many representations of networks at Cybergeography [21]. Pansiot and Grad [12] mapped paths to 5,000 destinations. The Mercator project at USC [10] tries to get a picture of the Internet at a given instance in time.

In terms of long-term mapping, k claffy and CAIDA are collecting a number of metrics from the Internet with skitter [19]. They have mapped the MBone, and collected path data to major web sites. We choose to map to each known network, preferring to map to everything that exists, rather than everything that is used (i.e. the web servers). Our goal is to discover every possible path, not just those in use.

Internet maps are often laid out on the globe or other physical map. The desire to map the Internet to geography is compelling, but it tends to end up with dense blobs of ink on North America, Europe, and other well-connected regions3. However, connections to distant and more sparsely connected regions can be represented nicely, c.f. Quarterman's map of connections to South America.

The problem with this method is the well-connected areas remain thoroughly inked, without a prayer of tracing paths through them. One approach is to simplify the map, showing connections by autonomous systems rather than individual routers. This is akin to showing the interstates on one map, and then creating local maps for each state. However, the AS connectivity graph is, proportionally, more connected than the IP graph, so the graph is still not very legible.

Interactive visualization tools can aid in navigating a database like ours. One can zoom, query, and browse at will. It is hard to see the entire net clearly on a screen: there are far too few pixels. However, H3Viewer [11] [17] is one tool that looks like a good start to such a tool. It displays a spanning tree of the graph and allows the user both to navigate the tree and also view the non-tree edges.

next up previous
Next: Future Work Up: Mapping and Visualizing the Previous: Watching Networks Under Duress
Hal Burch 2000-04-18