There are many folks making the argument that the country doesn’t have enough carrier exchange points. An exchange point is a physical location where multiple carriers meet for purposes of exchanging traffic.
The idea of exchange points started in the late 1990s as a result of the Telecommunications Act of 1996. That new law allowed for voice competition, and the national voice network was controlled by regulated telephone monopolies. Big telcos like Verizon and the other Baby Bells resisted the idea of competition and made it challenging and expensive to connect to the public switched telephone network at their voice switches. This prompted the formation of CLEC hotels, which were essentially data centers near the large telco switches where competitive carriers could make connections with each other.
Over time, broadband grew to be far more important than voice, and became the focus for the desire to interconnect. Over time these same locations were renamed as carrier hotels or Internet Exchange Points (IXPs). The exchange points attracted last-mile ISPs, cellular carriers, middle-mile and long-haul fiber providers, corporate networks, and eventually, content delivery networks (CDNs) from companies like Google and Facebook. The CDN category has grown to include those who are exchanging AI traffic.
I don’t know if this is an accurate statistic, but I’ve heard several people claim there are fourteen or fifteen states that don’t have a major exchange point. This might be true if you use the definition of an exchange point to be a place where everybody meets. The major exchange points are in or near to major cities and include places like 60 Hudson Street in New York City and the Internet exchange point in Ashburn, Virginia.
The major exchange points only make sense for large carriers that are exchanging large amounts of traffic. I recently saw that in the 60 Hudson IXP in New York City that it costs $1,000 per month just to carrier traffic from one floor to the next inside the facility. Only carriers with economies of scale can justify the high cost of locating at the major IXP sites. Most IXPs are owned by somebody other than carriers and ISPs and are viewed as a real estate investment.
There is a big downside to ISPs that are not located close to an IXP. They must lease fiber transport with somebody that can carry their traffic from the local market to the nearest IXP. Such transport can be expensive, particularly for small ISPs who must buy such transport from the telcos or cable companies they are competing against. Lack of affordable connection to the Internet is one of the factors that stop ISPs from creating connections in remote rural areas far off the beaten path. It’s interesting that the BEAD grants want to fund the last-mile networks in such remote places but are not willing to find any significant middle-mile fiber needed to make the connection to affordable transport.
There are good arguments to be made for the creation of smaller IXPs that are closer to where ISPs operate last-mile networks. One benefit of more points is peering, which is the process of routing traffic in a way that avoids having to send it through the major Internet IXPs. An ISP might create a peering point to directly connect to somebody like Google or Facebook to exchange traffic. Peering saves on transport costs, but it also saves on paying a fee to exchange traffic at the big IXPs, which is charged per megabit. For most ISPs, more than half of their data traffic goes to and from a handful of companies like Google, Meta, Amazon, Netflix, and Microsoft.
Perhaps the biggest benefit for ISPs to meet other carriers close to home is the ability to create redundancy. ISPs have learned that it’s dangerous to send all Internet traffic to a single connection point. A fiber cut or electronics failure along that route means the ISP and its customers will go dead. Regional IXPs open up the possibility of routing traffic to multiple large IXPs.
You should study up on what Hunter Newby and Connect America are doing about this and why. The first deployment is with Wichita State University. Hunter is the guy who took 60 Hudson, the former Western Union building from the early 30’s, and essentially invented the IXP, and created many more around the country. There are 14 states with no IXP, but many other places with great distances, including crossing state lines, to the nearest IXP. You will learn why it is so important, and can be economical for ISPs of all sizes to have proximity it an IXP.
There’s a lot of shiny objects being floated around Internet Exchange Points. While it looks easy, there’s more to it than setting up a switch fabric and connecting. What Connected Nation is doing is little more than a land grab which can negatively impact real efforts to do good.
Unfortunately your terminology is not broadly utilized in the industry, and sadly plays into the FUD that both Connect America & Hunter Newby are spreading. Chris has a future show that will dive into all things IXP’s and interconnection, stay tuned.
There is a fundament difference between a “meet-me room” vs. an “internet exchange point”, the term “carrier exchange point” doesn’t mean anything. MMR’s are generally passive in nature with lots of fiber patch panels, often times managed by a datacenter operator (sometimes called a “carrier hotel” in media). MMR’s offer a buffet of providers to choose from.
An IXP is active deployment of switch(es) for parties to mutually run BGP routing between each other, generally over a single cross-x or physical cable. The vast majority of IXP’s within North America are non-profit in nature and operated by their local community, these two characteristics are crucial.
Typical hop counts for subscribers to most services is over a dozen. Even if they live in a metro area where you’d expect a short number of hops to a carrier. This is one of the primary arguments for more IXs. Get end users to be just a few hops away from an IX and then only a few hops to 99% of the services they use.
All discussions on buffer bloat and packet loss and reduced by reducing hop counts. fewer buffers to touch, fewer places along the path to get packets dropped. Likely less physical distance between end user and service.
I’ve made the argument more about opportinuty connections, ie, IXs geographically distributed so that ISPs are wannabe ISPs of all sizes can get access to 100G pipes. This is one of the primary limitations in rural areas, they basically wait for government money to a winner-first build and they are then stuck with a single provider with all the negatives that come with that.
More IXs, and IX in every county. micro/pico IXs that are purpose built to link back to >=2 primary IXs. Light weight sites with nothing more that redundant 100G layer3 switches and set rates for transport. Most of the fiber is actually already in the ground and just inaccessible to small providers. Large providers don’t want to participate but they should be forced to if they touch the ROW. in fact, ANY provider that touches the ROW should be forced to participate.
IMO, this completely eliminates any need for federal funding of broadband buildouts.