The Internet Mapping Project


The Internet Mapping Project was started at Bell Labs in the summer of 1998. Its goal is to acquire and save Internet topological data over a long period of time. This data has been used in the study of routing problems and changes, DDoS attacks, and graph theory.

In the fall of 2000, Ches and Hal moved to a spin-off from Lucent/Bell Labs named Lumeta Corporation. This company applies our topological discovery techniques to discover the perimeter of our clients' intranets. Both Hal and Ches have moved on from Lumeta.

I no longer runs the Internet Mapping Project. I am not sure what Lumeta is doing with it, if anything.

Introduction

This mapping consists of frequent traceroute-style path probes, one to each registered Internet entity. From this, we build a graph showing the paths to most of the nets on the Internet. We have no interest in the specific endpoints or network services on those endpoints, just the topology of the "center" of the Internet.

These paths change over time, as routes reconfigure and the Internet grows. We are preserving this data, and plan to run the scans for a long time. The database should help show how the Internet grows. We think we can even make a movie of this growth someday.

The simple layout algorithm produces some nice maps.

Maps

This data yields a large graph. It is not easy to lay out a tree with 100,000 nodes. Our programs jostle the nodes around according to half a dozen simple rules, simulating various springs and repelling forces.

We have made some maps from this layout. A map helps us visualize things, to pick out points of interest, and find things that warrant closer inspection. Once the layout is computed, the map can be colored to show a number of things. The Internet is its own space.

The layout can be colored in many ways: with geographical clues, network capacity, etc. An Internet atlas would be interesting. We currently have maps colored by distance from the test host, IP address, and geographic region.

These maps are quite smashing, if we do say so ourselves. The December 1998 issue of Wired Magazine has the layout generated from data collected in mid-September. Hal generated a color scheme based on the IP address of the nodes. This sick idea ("Excuse me, may I have a prettier Internet address please?") creates a color scheme that seems to match Wired's traditional typography. But it actually does show communities that share similar network addresses.

Here is a .gif of the layout appearing in Wired.

Where are you on the Wired map? Don't ask. With nearly 100,000 nodes on the map, an index would be a huge sea of small type.

Uses

This data has a number of uses, including collaborations with other Internet mapping projects (see below.)

The database is documented here.

There has been confusion about this database. It is not the picture itself, but the raw data of the traceroute paths. It is a compressed text file, not a Microsoft Excel, or other database file.

Mapping Details, or "What are you doing to my net?"

The net mapping program sends small UDP packets to random high-numbered ports, or ICMP ping packets, while varying the packet's time-to-live (TTL) field. The TTL is decremented on each hop out. When it hits zero, the death of the packet is reported back to the sender. We do not expect to reach a working host, much less an active UDP service.

The packets are sent with slowly-incrementing TTL fields. When a packet fails to return, perhaps because it was lost or dropped by a firewall, we try a couple more times, then give up, recording any return code.

Future Work

The early results looked like a peacock smashed into a windshield. Though you could pick out the major ISPs and some interesting details around the edges, the map was't very useful.

We now run the layout on the minimum distance spanning tree, and the results on a 36 inch plotter are very close to a nice map.

Here's a movie (13MB) of one recent layout.

This data cries for interactive visualization tools. We've tried 3D, which didn't help as much as you might think. Lumeta now has the Mapviewer product which allows our clients to dig down into these graphs and extract a great deal of data about routers and connectivity.

One goal is to collect the data over time, and make a time-lapse movie of the growth of the Internet. Time-lapses of the annealing process are already interesting: it writhes and squirms and such.

Links to other Internet mapping efforts