Transcript: Community Broadband Bits Episode 96

Thanks to Jeff Hoel for providing the transcript for episode 96 of the Community Broadband Bits podcast with Scott Bradner on network peering. Listen to this episode here.



Scott Bradner:  And so, you either get money from Peter or Paul or both.  But the money has to come from someplace to build up the infrastructure.


Lisa Gonzalez:  Hello.  You are listening again to the Community Broadband Bits Podcast, from the Institute for Local Self-Reliance.  This is Lisa Gonzalez.

Daniel, from Bozeman, Montana, asked for a podcast on peering.  So we reached out to Scott Bradner.  In addition to many other accomplishments, Scott is a member of the Internet Engineering Task Force.  And he's a former columnist for Network World.  As Scott puts it, he's been in the biz, for many years.

Peering is the process by which two Internet networks connect and exchange traffic.  While we stream a movie or send pics to our family, we don't concern ourselves with the business arrangements between ISPs and carriers.  Maybe we should.  Or we should at least have a basic understanding of how peering works, because it does affect our use of the Internet.  As federal officials discuss the Time Warner [Cable] Comcast merger and network neutrality, the media is focusing more on peering.  While it's a good thing to make more people aware of the process, it's better to have people like Scott to offer an expert's opinion.

The discussion gets a little technical near the beginning, but if you shy away from tech talk, keep listening.  The interview gets deeper into how peering may or may not affect network neutrality -- and the potential merger.  Here are Chris and Scott.


Chris Mitchell:  Welcome to another edition of the Community Broadband Bits Podcast.  I'm Chris Mitchell.  And today, I'm speaking Scott Bradner, someone who has been working on the Internet for many years.  Welcome to the show.


Scott Bradner:  Thank you.  Glad to be here.


Chris:  You've been a long-time member of the Internet Engineering Task Force of the Internet Society.  I got to know you, I think, because of your role of Internet governance, and some of the people that you've known for a long time.  And you've also been a columnist for Network World.  So I'm very excited to talk about peering with you today.  But maybe you can explain to people in the audience just how they can know that you know what we're talking about.


Scott:  I've been in this business quite a while.  I consulted on UUNet's peering requirements back in the early '90s.  And helped run a local network in the Boston area, NEARnet, for a number of years -- helped found and run it.  So I've been in the biz.  But you have to take my word for that.


Chris:  I think that as we get into it, people will quickly see your expertise.  I would just like to cite two things.  One is, your beard of Gandalfian proportions.  And the other is an incredible acronym.  In an industry, particularly in the early years, when people were known by their three-letter acronyms, you are SOB -- which I really admire.


Scott:  Well, thank you.  My mother doesn't, but I do.


Chris:  Well, it was her decision, I guess.  So...


Scott.  Ah, no, actually they didn't give me a middle initial, but when I started to use it, she objected.


Chris:  OK.  Well, you know, it confuses me, because I have to catch myself from calling you Scott O'Bradner.


Scott:  You know, that's more Irish than I am.


Chris:  All right.  OK.  So, the first question that I want to get into, at the basic level: what does it mean to "peer"?  How does peering actually make the Internet work?


Scott:  Well, peering is at two different levels.  And it's unfortunate that the same term is used at the two different levels.  One it the actual technical interaction between routers.  So, a router in one Internet service provider talking to a router in another Internet service provider, or to a router in their customer.  And that's BGP peering.  The Border Gateway Protocol [BGP] is the routing protocol.  And at that level, the routers are exchanging reachability information.

But what's more generally talked about when talking about peering and the Internet governance, or network neutrality space, is a business decision between two Internet service providers on how they're going to interact at the business level.  So, one Internet service provider talking to another one.  A small Internet service provider getting interconnectivity from a big Internet service provider is normally just a customer of that second ISP -- and will pay for that interconnection.

But if you have two Internet service providers of approximately equal size, and approximately equal traffic, one to the other, the business decision can be made to exchange traffic without cost.  So that the first Internet service provider buys one or more wires to the second one, and runs those; and then the second ISP buys one or more wires to the first one, and runs those.  So each ISP is running connections to the other ISP.  And they don't exchange money for the traffic that's exchanged.


Chris:  So I think it may simplify it a little bit to just maybe pretend that you and I are each an ISP.  And what you're saying is that, in a number of cases, if you and I are of similar size, we would just exchange data between our two networks, and not bother counting the bits.  Because it's easier to do it that way.


Scott:  That, and the fact that I would buy a wire to you, and run it, and you would buy a wire to me, and run it.  So we're not even including the cost of the links.  We're each responsible for our own links.


Chris:  OK.  And so, historically, we've seen that among similarly-sized providers.  And, to some extent, I think, those who have an ideological predisposition to not charging for this interconnection.  Am I right by that?


Scott:  Ideology generally doesn't come into it.  It's really similar -- it's a cost-saving thing.  There's no particular reason to exchange bills when it's going to be a relatively equal balance -- you have to get rid of the accounting charges.  And it's also truly a business decision.  It's not a moral decision, or a regulatory decision.  It's a business decision.  And it allows one ISP, where the balance may not be exactly even to decide that there's business reasons to peer -- to do cost-sharing peering -- even through the traffic isn't exactly balanced.  Maybe there's a particularly attractive website on the second ISP, and therefore the customers of the first ISP would benefit from that, and make the first ISP more attractive.


Chris:  With -- Google, I think, famously, has a fairly open peering policy.  And so their business decision is, I think, not predicated on an equal exchange of bits, but some other factor, then, presumably.


Scott:  Theirs, certainly, because they're the supplier of bits.  It's usually the receiver of bits that makes the business decision, because they're getting a lot of traffic that is going to load down their backbone.  And they've got to see a reason to do that rather than charge for it.

So, certainly, it would be in Google's interest for everybody to peer with it because then they would be able to push a lot of traffic onto other ISPs without having to worry about paying for it.  The money flows based on the traffic.  So if I'm sending -- if I'm an ISP, and you're an ISP, and I send you a lot of traffic, you're going to want money from me to deliver that traffic.

Where cost-sharing peering comes in, whereas I would send you a bill for the traffic you're sending to me, you would send me a bill for the traffic I'm sending to you.  But if they're about even, then the bills would cancel each other out.


Chris:  Right.  Now, the question that I'm struggling with -- and I'm not trying to tell you how I think it should be, I'm trying to get a sense of how to reconcile the way that I think a lot of us think about it, which is -- and the real problem here is video.  That's why I'm focusing on YouTube.  Because I think YouTube has really changed the amounts of traffic quite significantly.  But...


Scott:  Yes, it has.


Chris:  ... I think of it as -- a small ISP has customers that are pulling traffic onto it.  And Google or Netflix could simply say, look, we're just providing what your people want.


Scott:  All customers of an ISP are expected to pay.  Harvard has a connection into an ISP.  I have a connection into an ISP.  I pay for my connection.  Harvard pays for its connection.  Google pays for its connection.  And the size of the connection is commensurate with the amount of traffic we're going to send or receive.  All customers do that.  So that the ISP, when they figure out their budget, says, oh, I get money from customer 1, I get money from customer 2.  The fact that they're exchanging traffic is fine.  But I get money from both of them.  If I'm only going to get money from one-half of the conversation, then I'm going to have to charge that one-half more.

So when I -- when Google and YouTube connect to an ISP, that ISP expects to get paid by Google for its connection -- as it expects to get paid by me for my connection.  If the ISP was not charging Google, then it would have to charge me more for my connection, in order to make up the difference.


Chris:  Let me step back for a second, to say that what I see -- who's been working with municipal ISPs -- is this sense that a smaller ISP has to pay a third provider -- let's say Level 3 in this case -- for transit, to get access to most of the rest of the Internet.


Scott:  A small ISP will pay a large ISP, as their customer.


Chris:  Right.  And so, as a customer of that ISP, when I use Netflix, I'm costing money to my small ISP, because it has to pay to get that Netflix content.  And, if I'm a member of Comcast -- if I'm part of Comcast -- now, it seems like when I use Netflix, I'm actually helping Comcast to make money, because Comcast is getting paid.  And this is where, if you can clarify where I'm wrong in that, that would be helpful.


Scott:  Comcast is assuming that all customers will pay.  All of its customers will pay, no matter whether they're the source of traffic or the destination of traffic.  So, I'm at home -- a Comcast customer -- I pay per month for my Comcast connection.  All of the other Comcast customers, including Netflix, pay for their connection.  And Comcast transports the traffic amongst them.


Chris:  The effect seems to be, though, that Netflix works out very well for Comcast, and very poorly for small ISPs.


Scott:  I agree that that's a significant problem.  In general, the question of how to deal with heavy traffic -- for the people who are cutting the cord with cable companies, for example, which is this particular case -- is a very serious issue.  And it's not clear how to fix that.  You make it so that a small ISP must accept large amounts of traffic from something like Netflix without charging for it, means that -- without charging Netflix for it -- means that they would have to charge their customers for it.  Either way, they've got to build up their infrastructure to support the traffic.  And so, you either get the money from Peter or Paul or both.  But the money has to come from someplace to build up the infrastructure.


Chris:  Now, as we're discussing this, I think, you know, what I've been advancing is that -- is what we've heard from many of the consumer advocates, who are very concerned that Comcast is abusing what is effectively a monopoly position.  And what I hear you saying is that this is the way networks have always worked.  Now, is it -- would you also be concerned with the potential of an ISP to abuse its monopoly position, to try and take advantage of this situation?


Scott:  Yes and no.  It is the case that in the U.S., that in almost all circumstances, the ISP for residential service -- in some particular geographic service area -- is a monopoly.  That's a fact of life.  That's not a particularly nice fact of life, but you don't have two high-speed Internet connectivity suppliers in most environments.  There are a few places where you have FiOS and Comcast, or FiOS and Time Warner, or one of the other companies at the same time.  But that's relatively few.  In most of the U.S., there's only one real, heavy-duty broadband supplier.  And that's a real problem.  We don't have competition.  Competition is what drives prices down and services up.  And we don't have competition.  That's a real problem.


Chris:  So, if I was to put it maybe a little less charitably, might you say, hey, you people that are not worried about there not being a choice, and monopoly, don't mess around with the connection -- that's worked well forever.  Let's solve the problem in a different way.


Scott:  Interconnection by itself won't solve anything.  The fact that you're -- if Comcast was told they must take all the traffic from Netflix for free, that would just give them more excuse to raise the rates to the individuals.  It doesn't solve the fact there's no competition for Comcast in most of the marketplace.  And certainly, the proposed merger with Time Warner means that the marketplace is still larger.  It's annoying that that merger will likely not be stopped for anti-trust issues because Comcast doesn't compete with Time Warner anywhere.  And that would be a requirement for anti-trust blocking of it.  The fact that it turns Comcast into a mega-ISP, covering most -- a good chunk of the U.S., is not relevant to that particular question legally.  So, it is not interconnect where the problem is.


Chris:  The way you phrase it, actually, is something that I've been playing with in my mind a little bit.  Which is that, that it seems like the collusion of the cable companies, in their refusal to compete, has led to a defeat for anti-trust policy, which is -- cannot stop the merger.


Scott:  Yup.  We've had that for a while.  And that question has come up as long ago as -- more than a decade ago.  The peering thing is a red herring, but it's a potentially important one.  I don't know whether you remember that, back in the early '90s -- or mid-'90s -- AT&T built a very large, very expensive, nationwide backbone.  And then went to a number of the then-big existing ISPs and said, I want to peer.  And they all said, our peering policy is that we have to be exchanging approximately equivalent amounts of traffic.  And you, AT&T, don't have any customers yet.  So, of course, we're not going to peer.  To peer would be to subsidize you.  It would be reduce your costs at the expense of ours.  And so why should we do that?


Chris:  Is it a novel environment now, with so much asymmetric -- which is to say that so many of the -- so much of the traffic on the Internet is destined for a cable or DSL network, which simply does not allow for symmetric traffic -- does that raise any problems?


Scott:  It would raise a problem if I, as a Comcast -- a subscriber to a Comcast asymmetric service wanted to be a source.  It doesn't -- if I am mostly pulling down stuff, like watching TV, watching over-the-top TV, the fact that the Comcast link is asymmetric doesn't make any difference to me.  But if I were to provide a cable -- a video -- service out of my house, if the Comcast rules would let me do so, then the asymmetry would be a big difference.


Chris:  A deeper question might be, how could anyone exchange traffic with Comcast, or any cable provider, on roughly equal proportions?


Scott:  Level 3 was doing that, before they got Netflix as a customer.  Level 3 and Comcast were peering.  And they were giving -- cost-free peering.  The fact that the last mile is asymmetric really doesn't make any difference.  It's who else is providing you information on the net.  If all I've got is residential customers, then the asymmetry might make a difference.  But none of the big ISPs only have residential customers.


Chris:  We've heard this discussed as a network neutrality issue.  And some have said it is, some have said it isn't.  Where do you come down?


Scott:  It is not.  Network neutrality is differential handling of traffic from different suppliers.  If Comcast were preferring Netflix traffic to other traffic, that would be a network neutrality issue.  Here was a case where, for whatever reason -- and primarily the reason was that interconnects between Level 3, which was the supplier Netflix was using, and Comcast were not up to the very, very heavy load of Netflix traffic -- that Netflix found that it was more beneficial to its customers to directly peer with ISP that was providing its customers.  And it did that with Level 3, it's doing that with Comcast.  But it's not a network neutrality issue, because it's not a question of Comcast doing any preference for the Netflix traffic in any way, as far as anybody has been able to tell.


Chris:  When people -- when I've heard that discussed, I sort of shudder a little bit.  Because it seems to me that, as important as, I think, some of the issues behind network neutrality are, it's actually become the forest rather than a group of trees.  In that, I think, the current structure of the market presents a number of problems, one of which is network neutrality.  And instead of people saying, there's problem with the way we've structured this market, they're saying, everything's network neutrality.


Scott:  Well, network neutrality has -- done by those people who favor it -- I'm one who does -- is a compensation for the lack of competition.  If you had two ISPs in -- or five ISPs -- in the Cambridge area, where I am, and they competed on the basis of what services they were offering to their customers that would be cool.  But there is really one.  It's Comcast.  And there's no -- FiOS doesn't make it to Cambridge.  So there's really one, it's Comcast.  And they don't compete with anybody.  So if they -- if Comcast were to decide to prioritize traffic for one of its customers over another one, I wouldn't have any other place to go, to balance them out, to try to keep honest.  Now, Comcast, it happens, they signed an agreement when they went and purchased NBC Universal, that requires them to meet the FCC's network neutrality rules -- unlike most ISPs.  But there's no question it's a serious issue.  And the serious issue is, we don't have any competition.


Chris:  You've been involved with the Internet for a very long time.  And I'm curious if you would agree with my perception that there's a concern that, right now, we're moving from an era of many different autonomous systems to more centralized, where we have a few mega-providers on the Internet, rather than a whole lot of small ones.  And is an issue of concern if it's true?


Scott:  Well, it IS true.  But it's not new.  It's been happening since the -- since at least 2000.  We've been moving to more and more concentration in the residential space.  Not so much in the business space, in the commercial space.  But in the residential space, more and more consolidation, with fewer and fewer ISPs.  Fewer and fewer, bigger and bigger ISPs.  And ones which almost always were carriers -- either telephone or cable carriers, in the past.  So, moving away from companies whose business was being an Internet service provider, and therefore whose business was moving bits, to companies whose business has been content.  And it's a different philosophy.


Chris:  So, are we losing something?  I mean, is the Internet less "Internetty"?


Scott:  Not really.  There's still three or four thousand ISPs in the U.S.  Most of them are baby ISPs, little neighborhood things.  But most of the -- the vast percentage of the residential customers are picked up by the carriers.  And that's increasing dramatically with the move to wireless -- the move to smart phones and the like.  There are an awful lot of folks who do not have any wired connectivity at all.  They're doing it entirely wireless.  And fewer of the big wireless carriers.  So, Internet via wireless, Internet via cable TV, you're getting down to a very small number of providers.  And that certainly is a significant issue, particularly when some of those providers, like the cable companies, have an arm which is all content, and yet the customers want to move away from the direct content provisioning, and go over-the-top and get their content separately.  And that's a conflict of interest.


Chris:  Well, thank you so much for coming on the show, and helping us to understand one of the more complex aspects of Internet policy, and something that's been a bit mangled in the news.  We really appreciate it.


Scott:  Sure.


Lisa:  For more on peering, follow our tag at  Thanks again to Daniel in Bozeman for the idea for the show.  And we want to encourage you to send your ideas too.  Write to  You can follow us on Twitter.  Our handle is @communitynets.  This show was released on April 29, 2014.  Thanks again to the group Valley Lodge for their song, "Sweet Elizabeth," licensed using Creative Commons.  Have a great day.