What is an "Edge Provider?"

3/01/2015 01:42:00 PM
Tweetable
On Friday I posted my thoughts on the recent FCC decision to reclassify internet service providers as common carriers in order to protect net neutrality. You may have noticed my post didn't mention so-called "edge providers" which have been a key point of contention in the net neutrality debate. The reason for that is the term "edge provider" is completely bogus.

Let me explain.

Here's how the FCC defined terms back in their first attempt at net neutrality regulation in October 2014:
Edge Provider. Any individual or entity that provides any content, application, or service over the Internet, and any individual or entity that provides a device used for accessing any content, application, or service over the Internet.

End User. Any individual or entity that uses a broadband Internet access service.
They interpreted it to mean anyone sending packets over the internet. As the Washington Post noted, that includes literally everyone on the internet. To read this blog post, you had to send an http request to my blog server, which then replied with the html text, which then caused you to send off several other http requests back to my server to get image files, CSS files, javascript files, and more. If you look at that Taylor Rule Calculator in the upper right on this blog, you actually sent four different http requests to the St. Louis Federal Reserve servers to get that data. Congrats! The FCC says YOU are an edge provider!

The FCC is wrong of course. The term "edge provider" meant something to network programmers long before the FCC and the net neutrality debate corrupted the term. To understand what an edge provider is, you must first learn about the provider edge. I've made a crappy graphic to help:
There are two Internet Service Providers, Time Warner on the left, and Comcast on the right, which connect various end users, denoted by the thin spokes, to each other. When an end user on one provider network wants to communicate with an end user on a different network, that network traffic must cross the "provider edge"--a super-speed connection between ISPs, shown here as the thick line between Time Warner and Comcast. Providers must agree on how to split revenues in order to maintain the common provider edge infrastructure.
In this graphic, there are two Internet Service Providers (ISPs)--the ISP on the left is part of Time Warner, while the ISP on the right is part of Comcast. The small spokes coming off of each ISP are the connections to individual "end users"--that is, internet subscribers--of each respective ISP. The giant connection between Time Warner and Comcast is a super-speed internet connection between two (or more) ISPs known as the "provider edge."

Note that in the graphic, Netflix, as well as the two Netflix subscribers, are all end users. All three have paid their respective ISPs for access to the internet. Consider what happens when Netflix subscriber 1 streams videos online: Time Warner collects money from Netflix for uploading those packets, and then charges Netflix subscriber 1 again for downloading those same packets. Both Netflix and the subscriber pay Time Warner for that activity. Now consider the case where Netflix subscriber 2 watches Netflix videos: Comcast gets money from Netflix subscriber 2 for downloading those packets, but doesn't get any money from Netflix for uploading those packets--the portion Netlfix pays actually goes to Time Warner instead of Comcast. Thus, from Comcast's perspective, Netflix is an "edge provider" and not one of their "end users." On the flip side, from Time Warner's perspective, it's Netflix subscriber 2 that's the edge provider, sending all these http requests to one of their own end users (Netflix) without paying them for the internet connection.

But the distinction is moot. Yes, in the case where Netflix streams video across a provider edge, each provider only gets half of the revenue, but then they each did only half the work! The issue is not that edge providers are allowed to free-ride on each other's networks--everyone has paid for their internet access--but rather that Time Warner and Comcast need to negotiate agreements between themselves on how to split their revenues to maintain the provider edge infrastructure. And it's obvious why these two companies want to charge Netflix twice, rather than just negotiate between themselves: Time Warner and Comcast are now the same company!

Now let's talk about congestion. According to the internet service providers, they need to be able to price discriminate in order to prevent the provider edge from becoming congested. First, note that the physical cables are not the limiting factor on how fast data can go through the provider edge--rather, in almost all cases, the limiting factor is actually the provider edge router, which is essentially just a computer that sits at one point on the provider edge in order to act as a traffic cop, directing each of the incoming streams from "edge providers" to their respective destinations at "end users." So when we talk about "upgrading networks" we are almost always talking about replacing these routers, not digging up and replacing the miles of cable between them. As I noted yesterday, ISPs can congestion-price traffic over the provider edge without discriminating on the basis of content (with the caveat that as I mentioned ISPs can negotiate between themselves on how to divide that revenue). Economically, those congestion fees, whereby people pay more to send and receive packets at high-traffic times of day on high-traffic parts of the network, can fully cover the costs of upgrading this provider edge infrastructure to maintain the Pareto-efficient level of service.

My point in all this is that once you note that ISPs can negotiate with each-other over how to maintain the provider edge, the actual distinction becomes irrelevant--they will just negotiate to the same equilibrium as if all the end users subscribed to a single ISP. Hence, "edge provider" is a completely meaningless distinction, economically speaking. The anti-neutrality lobby has re-purposed a relatively mundane semantic distinction to confuse people into supporting their position which, conveniently, also allows them to extract enormous rents from internet users with no other options but to pay the ransom.