Home / Blogs

A Possible Missing Piece of Net Neutrality Puzzle: Backbones and Peering?

I remember being told three years ago that, in general, internet backbone issues weren’t really a subject for regulatory involvement, and didn’t need to be. Although the last mile was a problem, the upstream fat-pipe relationships weren’t - they were all competitive and thriving. Or at least that’s what people thought.

Over the last couple of days I’ve been looking around trying to figure out what the facts are about backbones and peering. It seems that we don’t even know what we don’t know (“we” being the public). It’s an interesting area. Wikipedia has a good article on peering, but I can’t find a visualization of data (or even the data itself).

CAIDA makes clear, via kc claffy, that data about what happens on backbones is not available to us or, more importantly, to researchers. Gordon Cook says that everything about prices for backbone carriage is secret.

Why does this matter? Perhaps this is too simple, but if large ISPs (including traditional incumbent telephone companies, here or in other countries) have the market power to refuse to carry the traffic of smaller/competitive ISPs, or to condition the carriage of this traffic on agreement to particular discriminatory policies, then the neutrality problem just goes up a level. It ceases to be a “last mile” problem and becomes a backbone problem. If all arrangements carried out by large carriers are private and secret, then there isn’t even a platform for a policy discussion - traffic carried by (say) nondiscriminatory, smaller ISPs will just go more slowly.

In a way, the backbone issue (if there is one) potentially bears the same relationship to “network neutrality” that government ownership/control of spectrum bears to our current scuffles over spectrum policy: we may be missing an enormous part of the issue without knowing it. It’s as if we’re trying to describe the “issue,” the small vessel of points and counterpoints, without seeing that the vessel is housed in a gigantic, fortress-like, and mostly secret building.

(In America, as in many other places, our government controls a huge amount of spectrum without paying for it or even carrying its value on its books. Or even making precisely clear how much it really controls.)

So maybe I’m misunderstanding the importance of this issue, or maybe there really is a competitive backbone marketplace out there. But I wish there was more information about this. It seems fundamental.

By Susan Crawford, Professor, Cardozo Law School in New York City

Filed Under


Tom Vest  –  Sep 12, 2007 6:40 PM

Hi Susan,

The idea of juxtaposing peering with net neutrality issues is certainly intriguing, but I think that a great deal more clarity and precision of definition needs to be applied to both topics before they can usefully inform each other. To that end…

1. Peering itself: is what happens between two mutually “independent” “networks” (technically, “routing domains”) when they physically interconnect, in principle regardless of the commercial terms governing that interconnection.  The most common kind of peering is usually called “IP transit”, and involves one party paying the other in return for mediating all IP traffic exchange with (i.e., providing a “default route” to) the rest of the online world. In popular parlance, “peering” is usually used to denote those interconnections that have been established to exchange bilateral traffic only (including traffic generated by IP transit customers) on a reciprocal settlement-free basis.

2. Who can peer? Officially, to come into being as “independent” in the above sense, each network has to enjoy administrative control over at least two such physical connections, to two other equally independent networks. Note that both the question of which networks possess the technical quality of “independence” and which possess the means to peer (or to “peer”) are all fundamentally tied to the physical means of interconnection—i.e., who can obtain them (e.g., licensed telcos, anyone with a router, etc.), in what form (e.g., passive in-building cross-connects, “private line” circuits, optical wavelengths, dark fiber, right-of-way to construct a new path, etc.), under what terms (short-term lease, capital lease, IRU, fee-simple sale, DIY greenfield construction), for what purposes, etc (commercial reuse, R&E only, etc.). Arguably the fully loaded price of such links is the single most important variable explaining many of the Internet’s ostensible mysteries—why it emerged at the time/place it did, why and how it developed from a campus and inter-campus experiment into an ever-expanding and diversifying global economic sector, why market structures and concentration levels vary so much across countries, why it grows rapidly in some places and slowly or not at all in others, why some regions are perennial Internet “exporters” while others remain dependent on network services “imported” from external providers, etc.

Today there are over 26,000 active networks that satisfy the “independence” requirement. That means that, at the network service provider level, there are some 67+ million theoretically possible peering connections. Of course, the Internet is designed precisely so that direct interconnections are not necessary in order to exchange traffic—and in any case most of the 26k networks would rarely if ever experience any need to exchange traffic packet with the most of their counterparts. This is because the majority are not ISPs or online content providers in any conventional sense, both rather “network enterprises”, i.e., institutions for which the Internet is a critical input, but not a primary or public output. Even among the minority that do provide Internet access or online content as a public service, most experience high demand for traffic exchange—i.e., their respective customers are frequent senders and/or receivers of traffic—with only a small share of the others, for purely mundane conventional reasons (mutual relevance, linguistic, geographic, and cultural proximity, etc.).

To be continued…

Tom Vest  –  Sep 12, 2007 6:42 PM

Continuation of comment #1

3. Who can (and does) “peer”—and why? Notwithstanding the apparently contrasting definitions above (1), settlement -free “peering” is really just a special case of IP transit, one which generally has relevance only under a couple of special (and fairly rare) circumstances. Although there are a few ISPs that advertise and maintain an “open peering policy”—meaning that they will peer with anyone (or rather, anyone that can invest the resources required to extend their own network presence to a peering point designated by the “target peer”), most settlement-free peering arrangements are not really free at all, but rather in-kind exchanges which are perceived by both parties to be approximately equal in value. To clarify, network operators have traditionally used two rule-of-thumb metrics to estimate the relative benefits of interconnecting with each other. The first focuses on network investments—how many points of presence, spread over what geographic area, interconnected at what capacity, etc. This provides (or used to provide) a very very rough benchmark for a network operator’s capital investment commitment, and thus his/her total long run direct cost of service delivery. The second metric focused on volume and balance of bilateral traffic exchange—i.e., how many (x) bits per second exchanged, in both directions, consistently, measured over time at each point of interconnection. This metric helps to quantify the relative *benefit* of interconnection for the would-be peers, because it suggests how much less IP transit (and hence, short-term operating expenditures) would be needed if a new settlement-free “peering” relationship were established. The important thing to note is the conditions under which “peering” or settlement-free interconnection are generally agreed and sustainable over the long-term:

—Scope of network investments are “equivalent”
—Volume of bilateral traffic exchange is “equivalent”
—Balance of bilateral traffic exchange is “equivalent”

IFF these conditions apply, then it is theoretically possible for two operators to interconnect without having to come to agreement over the specific direction, unit, or rate of settlement payments. This does not imply that such values cannot exist, but simply that it’s possible to skip the intractable negotiations that would be required to figure them out. In effect, “peering” evolved as a convenience, back when metering traffic exchange was too difficult and/or expensive to warrant the effort. It persists today only in cases where operators are completely indifferent to concerns about asymmetrical costs and benefits, or when network operators have calculated that the value exchanged thereby is roughly equivalent. 

4. Who knows about “peering”? You are of course right that many facts about “peering” remain confidential. However, I wonder how different this is from other B2B (or intermediate supply chain) commercial exchanges in less prominent economic sectors, e.g., between steel producers and auto parts manufacturers? That’s not an argument in favor of confidentiality, just a question about whether the Internet is exceptional in this sense. In any case, since 2000-2001 many of the largest ISPs have openly published their settlement-free peering criteria (see examples below). Of course, as in other domains of inquiry, some “hardcore realist” observers of Internet development have dismissed such published policies as irrelevant; dominant operators will find an excuse to peer with you if they find it in their interest, they claim, and find an excuse to reject any peering request that they don’t favor, regardless of the published guidelines. As a former peering coordinator for a large ISP, my experience suggests that a more nuanced “institutionalist” view fits better. The published guidelines may not be the end of the story, but they do set common terms for negotiating interconnection that make it much easier for qualifying would-be peers to press their case, and much more difficult for peering-averse incumbents to arbitrarily reject the legitimate peering inquiries.

To be continued…

Tom Vest  –  Sep 12, 2007 6:45 PM

Continuation of comment #2

5. Selective “peering” a violation of net neutrality? I think your suggestion comes awfully close to the claims of some NN-skeptics (e.g., Yoo, PFF, et al.) to the effect that the Internet has never been “neutral”. I think this is an unfortunate understanding of the meaning of “neutrality,” and one that could tend to to diminish the novelty of some forms of market power that are emerging (or re-emerging, at least in the US), and the novelty of some forms of discrimination that may result. It seems unlikely to me that “peer” networks will find it in their interest to discriminate against some forms of traffic exchanged with their counterpart “peers”, because that would tend to undermine the interests of their own current paying customers, as well as reduce the cost-effectiveness of settlement-free interconnection itself. However, that’s not to say that in the future dominant networks might start to selectively charge their future paying customers based on mix of traffic (e.g., bundled search vs. third-party search), or decline to carry some kinds of third-party traffic altogether (e.g., third-party VoIP, etc.). This is unlikely to happen whenever/wherever competing, mutually independent, non-colluding Internet access providers are accessible by every would-be Internet user—but as the history of the Internet in the US and abroad makes plain, the existence of such competition is never assured.

6. Distinction between “last mile” vs. “backbone”: was once a reasonably accurate description—and remains a great idea, arguably, wherever it survives. In the US, virtually all of the surviving “last mile” operators now own their own “backbones”. Increasingly, they own everyone else’s as well. In places where the distinction still holds, “backbones” tend to be numerous, and tend to function as the mechanisms that enable others to “route around” unappealing anti-features like selective service degradations and other NN-contrary behavior. Generally speaking, it’s only where such territories cease to be addressable by such extra-territorial “backbone” networks—and no one’s left but the vertically integrated territorial incumbent—that you begin to observe the implementation of such measures beyond the last mile.  Thus, describing “peering” as a “backbone-level” threat to net neutrality may not be entirely wrong, but cases of such abuse are hard to find except where where the distinction between “last mile” vs. “backbone” itself does not exist.

Sample Peering Policies




Comment Title:

  Notify me of follow-up comments

We encourage you to post comments and engage in discussions that advance this post through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can report it using the link at the end of each comment. Views expressed in the comments do not represent those of CircleID. For more information on our comment policy, see Codes of Conduct.

CircleID Newsletter The Weekly Wrap

More and more professionals are choosing to publish critical posts on CircleID from all corners of the Internet industry. If you find it hard to keep up daily, consider subscribing to our weekly digest. We will provide you a convenient summary report once a week sent directly to your inbox. It's a quick and easy read.

I make a point of reading CircleID. There is no getting around the utility of knowing what thoughtful people are thinking and saying about our industry.

Co-designer of the TCP/IP Protocols & the Architecture of the Internet



Domain Names

Sponsored byVerisign


Sponsored byDNIB.com

Brand Protection

Sponsored byCSC

IPv4 Markets

Sponsored byIPv4.Global

New TLDs

Sponsored byRadix


Sponsored byVerisign

Threat Intelligence

Sponsored byWhoisXML API