Internet Exchange density in the United States

Yet another PeeringDB map

I’ve heard some rumblings about a crisis of availability of Internet Exchanges in the smaller markets in the US. I wondered, is this a real problem?

When the IX fanatics, myself included, talk about the benefits of connecting to an IX one of the primary advantages given is keeping traffic local. This is also referred to as reducing latency in network engineer speak. So are there large parts of the country that don’t have nearby access to an IX?

Well, here’s a rough map.

The lines show the closest IX fabric and are color coded as following:

• Green- 0-160 km (~2 Optical hops)

• Yellow 161-320 km (~2-4 Optical hops)

• Red 321+ km

The lines run from the center of each county in the US back to the nearest PeeringDB facility that houses an IX fabric.

Here’s some my initial thoughts:

• Like many of these maps, if you zoom out a bit it looks like a population density map. This makes sense as areas with more population would be able to support a larger number of IXes.

• The largest red areas (non-Alaska) are the very sparsely populated great plains. This area has a very sparse population but the counties are smaller than the counties in the mountains further west.

• The western mountain region has surprisingly good coverage with multiple community-driven IXes.

• Alaska looks terrible. Historically, there have only been a handful of ISPs in Alaska and they seem to have agreed to exchange all their traffic at SiX in Seattle. There is a project underway to start and IX in Southern Alaska. I also think that Starlink might help that along.

• The line into Canada appears to be a data fill issue.

Thanks to SimpleMaps.com for the US county data!