18 Comments

Summary:

The interconnection agreement Comcast and Netflix signed isn’t a violation of network neutrality, but it continues a troubling precedent for the internet and has anti-competitive overtones.

net-neutrality-scales

It wasn’t long after Comcast and Netflix finally settled a festering peering dispute that people starting claiming that this resolution means the end of network neutrality. But it’s not. For better or worse, the legal framework governing network neutrality in the U.S. — the recently gutted Open Internet Order — doesn’t touch the issue of network interconnection and peering.

It could have, but at the time the FCC was pulling together the Open Internet Order in 2010, then-chairman Julius Genachowski decided not to bring interconnection agreements under the purview of the order. At the time it made sense, though even back then peering disputes raised consumer-protection issues. It would have precipitated a big fight and upset a fairly established internet practice when the big focus was on the last mile. So now we have a situation that FCC chairman Tom Wheeler summed up well when I asked him about peering in an interview late last month:

”A lot of people seem to think the whole peering and interconnection topic is the same as net neutrality. It’s not, it’s a different issue — it’s a cousin, maybe a sibling, but it is not the same issue. But it is an issue that is something the commission has to stay on top of. Our job is to make sure whatever happens is not anticompetitive, is not favoring one party…and that’s the challenge we have to apply to make sure it is a competitive, vibrant, non-preferential market.”

So what’s going on?

Even if Netflix (and other companies) paying Comcast for direct access to its network isn’t a violation of network neutrality, it is a concern. The fear is that Netflix can’t provide a decent-quality video streaming service unless it pays to peer with Comcast, either directly or through a transit provider that has a direct interconnection. Netflix has mitigated this issue by paying transit providers that peer directly with Comcast to carry its traffic. But as anyone who covers the internet knows, internet giants benefit if they can control their own infrastructure and costs as they get bigger.

With 30 percent of U.S. broadband traffic, Netflix is certainly big enough to want to control the costs of delivering its bits to the end consumer. It tried to do this in 2010 by shifting its content delivery network agreements from Akamai to Level 3, which would act as both a CDN and a transit provider. The idea was that Level 3 already had a settlement-free (it didn’t pay) peering agreement with Comcast, and Netflix could benefit from that relationship without paying Comcast directly. After the influx of Netflix traffic upset the balance that Comcast and Level 3 had, Comcast and Level 3 engaged in a very public peering fight.

Level 3 agreed to start paying Comcast. In 2012, Netflix tried once again to control its own destiny with the introduction of Open Connect, a content delivery network that offered ISPs servers that would cache the Netflix content. The larger ISPs didn’t bite. So with its capitulation that I reported on Friday, Netflix is accepting the idea that unless it pays Comcast directly or through a transit, it can’t deliver the user experience its customers expect.

A modest proposal

Given that Microsoft, Google and many others have already decided to pay Comcast for a direct connection or pay a transit provider with a direct connection, Comcast has now enshrined its version of peering — one that requires companies that send a lot of traffic to its network to pay. An argument can be made that such payments help Comcast expand its network to meet the demand that services like Netflix put on the infrastructure, although the counter-argument is that the end consumer paying for broadband is footing that cost.

I view it as ISPs figuring out how to impose the old “sender pays” model on the internet — something we at Gigaom view with no shortage of alarm.

Another issue is that Comcast is setting the rates to let content providers get their traffic onto its network rates in a relative vacuum. These agreements aren’t transparent, even within the industry. Do these rates match the cost of buying additional ports or wavelengths to carry the influx of traffic? Could Netflix figure out a way to do this more cheaply if left alone? Could Google? How much can Netflix or another provider control about the type of gear used or where the networks meet?

These rates could go up over time, and they essentially act as a tax on the internet that the web content companies have to pay to ensure their service gets to the end user. They pay this tax directly if they are a big enough provider, or indirectly by locating their servers in places that peer directly with Comcast or buying transit.

For an internet that has been built on the concept of settlement-free peering agreements that don’t require a CEO-level negotiation, this seems like a worrisome development. It could add bureaucracy and cost and enforce the ISP as a gatekeeper. Luckily, there’s a pretty clear solution. Just open these contracts up to industry and public scrutiny. Perhaps Chairman Wheeler will get on that.

comcast

  1. It’s hard to weigh all of the intricacies of what is obviously an extremely complicated issue, but both the bigger issue and this sort of specific agreement seem overtly problematic. Not only do they raise issues about an open internet, but they also put those who can (afford to) pay (the likes of Netflix, Google, Microsoft, Facebook, etc.) at a distinct advantage over existing and future competitors who will have new barriers to complete with. The fact that consumers are asked to pay more and more for the same utterly mediocre service is bad enough…

    Share
    1. Can you name three services that create 10% as much traffic as Netflix and can’t afford a direct peering agreement?

      Share
      1. Intentional straw man, or…? I started by suggesting the issue is complicated. An obvious source of complication would be the fact that Netflix is generating a ton of traffic, right? Another obvious source of complication would be that consumers pay ISPs for access to “the web” (not “the web” with throttled Netflix), right? Of course, neither of those issues make the remaining fact that the agreement is problematic for the reasons I cited.

        Share
  2. I mean, the thing that everyone needs to remember is that these service DO cost money. Netflix is sending massive massive amounts of data. Somebody somewhere needs to pick up the tab for that. That’s not an insane outlandish concept.

    Share
  3. I have a question about this phrase: “For an internet that has been built on the concept of settlement-free peering agreements that don’t require a CEO-level negotiation…”.

    Which Internet is that? It’s not the one I use. The idea that one company can generate a third of the traffic load on the entire North American Internet should not have to coordinate with others and yes, pay some money for the facilities it uses, is a bit ridiculous. I know “innovation without permission” is the latest meme at the cool kids table, but it’s not really a sensible idea.

    Share
    1. It’s based on the packet-clearing house and OECD data that says 99.5 percent of peering agreements are done on a settlement free basis. http://oecdinsights.org/2012/10/22/internet-traffic-exchange-2-billion-users-and-its-done-on-a-handshake/

      As for paying money for the facilities it uses, I already pay for them as a customer of the ISP. This is the settlement model that allowed the telcos to control our phone and eventually broadband for decades, stifling innovation. As a consumer, I’d rather sit at the “cool kids table.” It’s not like Comcast is in danger of folding because it’s upgrading a few ports to handle the traffic that entices people to upgrade their broadband.

      Share
      1. The old peering arrangement was for the mutual benefit of network providers. Free makes sense assuming the networks generate roughly the same amount of traffic. Netflix is not a network; it is a customer of networks.

        Share
      2. Settlement-free Peering always assumed symmetrical traffic. Most agreements I’ve read always had a provision for compensation if the traffic was/ or became substantially asymmetrical.

        I believe these agreements are at the root of why CDN’s first began paying.

        In the future, Comcast could decide to extend xfinity to any broadband subscriber or Verizon could extend FiOS on demand to any broadband subscriber. A significant take rate would drive a spike in traffic that theses providers would have to address commercially with other broadband companies.

        When you net this out, the network is going to be optimized for video delivery. This is why Netflix is the test case. They are the biggest video streaming service and generate the most traffic. If we want our broadband providers to continue to invest in capacity expansion and speed increases, someone pays for it or they don’t invest.

        If you are a heavy consumer of video content and are looking forward to 4K video streaming, do you really want your broadband provider to be using the same interconnection they had in place when cat videos on youtube were viewable on the desktop PC?

        I’m totally for the ability of a consumer to access any legal content from any source without fear that their broadband provider will impair the service experience. Even though the effect may be service impairment, lack of network investment to support a better customer experience isn’t necessarily a violation of net neutrality principles. However, if you’re expanding every aspect of your network, but refusing to expand capacity for a CDN who is willing and able to pay for the infrastructure upgrade, because you’re trying to establish a competing service, then I think the regulators will have to step in. Parsimony and bad customer service isn’t generally illegal. Preferential treatment of traffic for competitive reasons could be.

        When you consider the combined Comcast, T, VZ, CLINK will own 75% of fixed broadband, and that 4 companies seem like they will share 70% of the LTE subscribers, then it is very hard to argue that there isn’t a market power issue in access. If these companies are caught violating NN goals, they’re inviting re-classification. Their worst nightmare.

        Share
      3. The “OECD data” in Weller’s report consists of conversations he had with a couple of guys named Bill. While it is very common for *similar networks* to peer with each other without a contract, this isn’t the way YouTube, Akamai, Amazon, Level 3, Limelight, of any other firm similar to Netflix connects to ISP networks so it’s irrelevant to the issue on the table.

        What is happening here is that Netflix currently connects to Comcast through Cogent at five peering centers. Since Netflix began streaming 4K, its traffic load exceeds the capacity of the ports between Cogent and Comcast. This deal will probably give Comcast 300 10G ports at ten peering centers around the country for roughly $5 – 15M a year (total). Given Netflix’ annual revenues are over $4.5B, this isn’t even a bump in the road.

        As for making new precedent, most people think this deal follows the template for YouTube and the ISPs. YouTube has some unique problems with interconnection because the ongoing spat between Google and Equinix. It turns out very few businesses are willing to give up something for nothing, but when the terms are reasonable it’s not hard to get reasonable people to play ball.

        Share
  4. Peering is for peers. ISP to ISP, CDN to CDN. It is not a universal way for large businesses to get a free internet connection. A large company like Boeing with thousands of employees and online content pays to connect to the internet because they are not an ISP. Comcast is an ISP. Netflix is not. If Netflix was smart it would start offering its customers names@netflix.net. Maybe partner with a mobile carrier. Peering might be an option then, not now. I pay for my business, why shouldn’t they? The internet is not just for consumers, it is big business.

    Share
  5. What if…

    Shell would charge GM for providing fuel to GM owners, because there are so many…

    It takes a lot of work to make that much fuel and many companies are involved in making the stuff and transporting it. You have to produce a lot of ads as well to tell those GM owners that you have the stuff they need. Not to mention the tens of thousands of people manning the gas stations, they have to sell the stuff. How can you expect Shell to make any money if it’s that much work. It is only natural that Shell charges GM for providing such labor intensive products. Some of those GM cars use so much fuel it’s really really hard to keep up. Now that you have me thinking of it. Those stupid farmers and their cows and other animals. All those stores need to service these annoying CUSTOMERS. It’s stupid. Why don’t these cows pay for the milk to have it sold at groceries.
    Actually I don’t even understand why we are having this discussion. It’s a free market, right? Where your product determines your success! It’s not like they are abusing their direct contact with these CUSTOMERS. This is the US not China, where the government and large institutions determine what you get as a CUSTOMER.

    Share
    1. Michael Salinger Monday, February 24, 2014

      @dirk robers,

      The problem is that this is *not* a free market, and that’s what makes this troublesome. Comcast, for all intents and purposes, is a monopoly. They have one goal in mind, maintaining as much profit as they can. Their customers are basically captive to them in most markets (in some there is one other choice, but a duopoly is typically not much more competitive than a monopoly).

      Comcast has no incentive to provide good service to their customers. Let’s consider a market where there were 5 or 6 different ISPs that a customer could choose from. Each ISP would have an incentive to attract customers. Let’s say company A determines that it can increase its customer acquisition substantially if it can demonstrably have better streaming quality for its customers. Company A then makes the decision to invest in its infrastructure (something Comcast has not been doing, and part of the reason for the problems, at least according to Level 3). Company A invests in this infrastructure in order to provide a better service and attract more customers – it takes on this cost itself because there is an advantage for itself to do so. When Company A makes traction in the market, the rest of the companies (to varying degrees) see what they did and the success that Company A had, and will mimic them or try to do better. The net result is consumers win and costs remain low or get even lower due to the competition.

      In broadband in the U.S. (which is *very* different than the rest of the world) there *is* no competition. The cable companies have a monopoly/duopoly in pretty much every market, and most of that is due to government regulations – i .e., this is a government-instituted monopoly. This results in the situation we are in now where there are *no* incentives for Comcast to improve its service, and they will basically gouge others (either customers or providers) to provide good QoS – because where else are they going to go?

      There are two solutions to this problem – deregulation to allow for competition, or tighter regulation to ensure that these monopolies/duopolies cannot abuse the power they’ve been given in the marketplace. I personally prefer the deregulation approach with increased competition, but that would require a rather large undertaking that may take years at this point. So, in the meantime, while we investigate and move towards a path of deregulation, we need an interim period of increased regulation to ensure that the monopoly power of the broadband providers isn’t abused.

      Share
  6. It is amusing and hypocritical for one internet based business (Gigaom) to criticize another Internet business (ISP) and suggesting government control of ISP but not businesses like Gigacom, Apple, or Google. Saying that consumers are impacted by the Netflix and Comcast deal is not much of an argument as that applies to everything dollar exchanged in the Internet. The advertising dollar that goes to Google presumably comes out in the end of the consumer pocket.

    If Comcast is a monopoly ISP due to its control of 30% of US market, so are Apple’s 40+% of smartphone and Google’s 60+% of search.

    Share
    1. Uh… Well, Comcast is literally the only competitive option for broadband internet in a large number of markets (mine, for example). Apple or Google may own major slices of the mobile device market, but they directly compete across the entire market. That’s an essential difference. I have many options for mobile devices and can change devices at any time. Most people don’t have any more than 1 or 2 options for competitive broadband internet service. If my choice of mobile device was Apple or nothing, the issue would be similarly problematic, but that isn’t the case.

      Share
    2. The issue isn’t Comcast’s market share across the entire country but the fact that its municipal franchises are de facto local monopolies. The cable industry has gotten fat and happy with this arrangement in which they rarely have to compete for customers.

      Share
  7. Net neutrality is an issue that needs to be fought for, and the first step is to keep up with the issue as much as possible. If anyone needs a refresher on the basic issues, here’s a great short mockumentary: http://www.theinternetmustgo.com/

    Share
  8. Christopher L Wood Friday, February 28, 2014

    Anyone who doesn’t understand the reason for this just doesn’t understand the big picture. Why do we think it is fair to have to pay a company like Comcast for more bandwidth, but unfair for Netflix to have to pay when they effectively use up more bandwidth as well?

    I see no slippery slope here. This is how it has always worked. Netflix users are only complaining because they can’t stream movies perfectly. This happens on every site though. You can’t always stream Youtube videos perfectly because of bandwidth limitations out there that have nothing to do with what you pay your ISP.

    If ISP’s did not throttle big websites with massive data transfer like Netflix, then all users would feel it. It would be no different than when one massive program starts hogging your system resources, and your whole computer hangs.

    But apparently some people out there are too idealistic about the very physical way that the internet works. It isn’t just a bunch of abstract software connections and “peering”… Can you imagine if software programmers programmed without any thought about hardware limitations?

    Share

Comments have been disabled for this post