The same internet route traces are rendered in two interactive cross-linked views: a geographic map on the left, and a logical connection-oriented graph on the right. Routes were traced from a single location located at the center of the graph view.
All traces originate from my computer in the Tampa, FL area, depicted as the red dot in the center of the graph. Moving outward from there, a dense core of my ISP's regional routers surround the center in a lattice-like topology. Remote destination nodes are typically located at or near the edge, but some endpoints appear closer to the core. The result of many individual traces were combined to yield this set of interconnections.
As this is a set of all observed interconnections, the actual path taken by a particular IP packet would traverse a limited set of nodes. This is particularly true near the center where many parallel interconnections exist for capacity and redundancy.
The graph's "Force" feature allows D3's force-directed graph layout to move nodes based on its constraints. While the initial version of the graph used this algorithm, you can try it to see that its result is not sufficiently clear and takes a long time to stabilize. This version uses a layout created in Gephi using Yifan Hu's Multilevel algorithm as it is better suited to a branching tree topology as exists here. It did create too many path crossings which I've manually cleaned in most cases.
This data is extremely limited since it was collected from a single location to a relatively small set of endpoints. Inaccuracies include omission of hops when routers failed to respond to a trace request as well as erroneous geocoding of IP addresses. My purpose is primarly to experiment with data visualization tools but it has provided some interesting glimpses into the fabric of the internet.
Here is a map presenting a schematic view of undersea cable routes. Unsurprisingly, it correlates strongly with the map on this page.