Tim Lee

Network Discrimination in the Real World

I’m trying to wrap my head around the scenario I criticized earlier today, where a broadband ISP charges individual sites for faster speeds.

Let’s suppose, for the sake of argument, that Comcast imposes a half-second delay into the loading of any website that doesn’t pay a special high-speed access fee. The fee might be $1 for every 100,000 page views. This website gets roughly 100,000 page views per month, so we’d owe about $1/month to Comcast if we wanted to avoid having our site load slowly for Comcast customers. A site like Techdirt, which gets roughly 100 times as much traffic as we do, would owe Comcast about $100/month if it didn’t want its traffic slowed. Google, which gets 100,000 times as much traffic as us, would have to pay about $100,000 per month. Clearly, such a scheme could bring in tens of millions of dollars in additional revenue each year.

Of course, it would be ridiculous for us to send Comcast a $1 check each month. Especially since we would presumably be expected to do the same thing with Verizon, AT&T, Charter, Sprint, Qwest, and dozens of smaller ISPs. Running a “high speed” web site would require writing dozens of checks to dozens of different network owners.


Of course, in practice, the process would be streamlined, probably by your ISP. So instead of having to send dozens of checks out yourself, you’d sign up with an ISP that has already negotiated blanket “high speed agreements” with the various ISPs on the Internet on behalf of all of its customers. The same would be true of turnkey hosting products like TypePad.

Now, it’s likely that the largest ISP would be able to negotiate bulk discounts. After all, if I’m representing 50,000 website owners, I’ve got a lot of leverage. If Comcast doesn’t agree with my price, it will be forced to degrade the quality of all 50,000 of my customers’ web sites. That will not only deprive Comcast of a lot of revenue, but it will also get Comcast’s own customers pissed off, because their Internet experience will suffer. Smaller ISPs, recognizing this, would probably ask their larger, upstream ISPs to negotiate on their behalf in order to give them greater leverage.

So the most likely outcome would be a small number of large business ISPs, each representing tens of of thousands of websites, negotiating with a small number of large residential ISPs, each representing tens of millions of consumers. Notice that the situation is largely symmetrical: if a residential ISP walks away from the table with a particular business ISP, it’s not obvious who is hurt more. The residential ISP will take just as much heat from its customers as the business ISP will take from theirs.

Here’s the thing: this isn’t a hypothetical situation at all. In fact, it precisely describes the negotiation of peering agreements among major Internet backbone providers. In those negotiations, each party would like to force the other to pay it for the right to interconnect. If there’s a big difference in size or degree of connectedness, the smaller and less-connected company pays the larger, better-connected one for a connection. If they’re the same size, they engage in a “settlement-free” peering agreement, in which each agrees to carry the other’s traffic free of charge.

In practice, the nine largest networks peer on a settlement-free basis, while the others pay them for access. I believe that five of those Tier 1 carriers (AT&T, Verizon, Qwest, AOL, NTT) serve mostly residential customers, while three others (Global Crossing, Level 3, and Savvis) serve mostly business customers. (I think Sprint serves a substantial number of each)

The rest of the Internet, directly or indirectly, pays one or more of those nine companies for connectivity. The prices for connectivity are set in an extremely competitive marketplace. Those backbone providers are already charging as much as they can get away with. It’s not obvious why those companies would suddenly be able to charge more when they label part of the bill a “high speed web access fee.”

Moreover, if the vast majority of web sites ended up having high speed agreements negotiated on their behalf by their upstream ISPs, it’s not obvious what point would be served by each residential ISP keeping a massive database of which web sites have paid for preferential treatment. Compiling, updating, and distributing such a database would be a massive administrative task. The database would have millions of entries, and it would have to be distributed to thousands of routers around the world.

The far more efficient structure is the structure that’s already evolved in the competitive conditions of the modern-day Internet. Individual ISPs don’t negotiate with individual websites because doing so would be an expensive and pointless duplication of effort. Large ISPs are already charging smaller ISPs for access to their “pipes” at market-determined rates. The system works great, and there just isn’t any reason to change it.

November 10, 2006 | Comments |

discussion by DISQUS

Add New Comment

Viewing 6 Comments

    • ^
    • v

    Tim, I think the bigger issue that you aren't wrapping your head around is that you're assuming that the network will always operate exactly as it does today. This is a 30 year old technology we're talking about that hasn't seen any real change to its basic operating protocols. To assume that the way it works now is the best possible way for it to continue working is woefully shortsighted, I think. End-to-end and TCP aren't written in stone, and from my understanding, the only reason they're so revered is because they are simply what is.

    I'm frankly pretty shocked that you're taking an if-it-ain't-broke-don't-fix-it approach to innovation, of all things.

    • ^
    • v
    Brooke:

    I don't see how your comment is a response to what Tim said, which is about contractual business arrangements, not protocols.

    Tim:

    I think you're exactly right--I fully expect to see established standards for inter-provider QoS and the negotiation of peering agreements which include terms for settlement on a per-class basis. Providers who see an equal value in exchanging traffic at the highest class of service will do so on a settlement-free basis; that may not (and probably won't be) identical to the current relationships providers for best-effort Internet traffic. A provider that is tier one for best-effort Internet may not be tier one for assured or enhanced levels of service.

    Your description of the type of customers that the top providers have is more-or-less accurate, but it is a little bit more complex than that, because there are lots of major consumer providers that purchase transit from providers in the "mostly serve business customers" category. For example, while Global Crossing has businesses as direct customers, those businesses include some very large consumer ISPs (mostly outside of the U.S.).

    It makes sense for customers to only pay their direct providers for higher classes of service--the situation of Google directly paying AT&T; for enhanced service would likely only occur where there is a direct customer relationship between them as well.
    • ^
    • v
    Well, I have no quarrel with his thoughts on contractual arrangements. I'm simply making the observation that as in his last post, he is assuming that everything in networking technology will remain as it is now. In the interest of being germane, I'll add that protocols might--and likely will--change, possibly rendering the discussion on contractual business arrangements as they exist today irrelevant in predicting future outcomes.

    Perhaps I'm too much of a non-expert in this area for my own good. Perhaps I'm having too easy a time imagining paradigmatic shifts in networking technology that seem outside the realm of possibility to someone more knowledgeable about the state of current technology.
    • ^
    • v
    Let's suppose, for the sake of argument, that Comcast imposes a half-second delay into the loading of any website that doesn't pay a special high-speed access fee. The fee might be $1 for every 100,000 page views. This website gets roughly 100,000 page views per month, so we'd owe about $1/month to Comcast if we wanted to avoid having our site load slowly for Comcast customers. A site like Techdirt, which gets roughly 100 times as much traffic as we do, would owe Comcast about $100/month if it didn't want its traffic slowed. Google, which gets 100,000 times as much traffic as us, would have to pay about $100,000 per month. Clearly, such a scheme could bring in tens of millions of dollars in additional revenue each year.

    What this will destroy is the dynamism of the web--now a start up won't have as much of a comparative advantage to the established playrs, as that start up will have to pay for their interconnection to the web. Google went from nothing to being a major player in just a couple of years, and created much value in the process.

    I think we need to think of the internet as basic infrastructure, and stop trying to erect toll boths.
    • ^
    • v
    Peerage is an interesting model to use Tim. However, you can't forget switching costs. It is not cheap or instantaneous to move from one provider to the next. And with a bit of lock-in comes leverage. Ergo, IMHO, something must be keeping the various tiers of ISPs from screwing each other in some way on peerage costs (and don't tell me it's ethics). It may be that they all own fragments of the network, so are all co-dependent, with little money actually changing hands. Perhaps the general legislation about predatory pricing & price fixing are enough to keep them honest too.

    So, is there enough already existing legislation to solve these predatory scenarios?:

    1. Consumer pays a mega premium because there's no other bband choice in their area.

    2. Content provider with a service that is directly competitive to a service offered by ISP is diminished in some way: latency, $, defamation, etc.
    • ^
    • v
    Enigma: I agree! I was simply offering it as a hypothetical to explore how it might play out. The point of my post (Which I may not have made very clearly) is that in practice, such a scheme wouldn't be profitable for ISPs.
blog comments powered by Disqus