The secrets of de-peering
Handbooks, conferences, best practice documents – there are many resources to turn to in order to learn about Internet peering. But when it comes to de-peering, networkers are mostly left with their gut feeling.
There is no “Beginner’s guide to de-peering”, no “De-peering 101”, no manual “How to de-peer respectfully in five easy steps”. This absence of guidance exists for several reasons. The most obvious being that it appears counter-intuitive to busy oneself with de-peering and cutting down transmission capacity between networks when traffic levels on the Internet have kept going up and to the right since its inception.
And indeed, the network engineers I have spoken to for my PhD confirm that de-peering is not common. However, when it does happen, it is instructive to look at; by way of the arguments networkers use to justify or assess de-peering, they offer a rare glimpse into the networking industry and the manufacturing of connectivity. It becomes explicit what are accepted conventions and where there run lines of conflict. For instance, it may come as a surprise for some that not every de-peering indicates a dispute; and conversely, despite press coverage not every “peering war” even implies de-peering.
Ordinary dictionaries do not (yet) include peering or de-peering, so let’s clarify the vocabulary. Networkers may want to skip this paragraph. Peering refers to a specific type of interconnection arrangement between network operators. On a technical level, to peer means to interconnect two Internet networks (autonomous systems in technical parlance) and exchange traffic that belongs to each other’s customers – be they end users or other networks. (So peering offers limited reachability, which is a key difference between peering and so-called transit. In transit, one network offers the other to transmit data to and from any endpoint on the Internet for money.) Economically, the dominant understanding of peering was for the longest time that did not involve the exchange of money between the parties who peered. It was settlement free. More recently though, so-called paid peering has surfaced as a variation. In that case one network operator pays for her customers to reach the other network’s customers directly. To de-peer then means to terminate a peering relationship and disconnect the networks.
Where networkers agree about de-peering
While de-peering always leads to loss of Internet connectivity between networks, it does not always cause a conflict between the networkers involved. In fact, there is a widely accepted rationale for de-peering among networkers: an industrial way of thinking. This rationale prioritizes functionality and efficiency. In the eyes of the colleagues, a networker may legitimately shut down a peering session when his network receives malicious traffic (such as a DDOS attack) from its neighbors, when it is sent incorrect route announcements, when it disconnects from an Internet exchange point all together, when it merges with another network or, when it experiences congestion due to a peer not providing the throughput capacity that had been promised. Networkers widely agree that protecting a network’s functionality and performance has priority over any other criteria for assessing Internet interconnection. In the words of one interviewee: “Someone spamming my network is a damn good reason to de-peer. Is that a net neutrality issue? Of course not. It is an Internet peering issue.” Often times, networkers re-establish peerings after such issues have been resolved, sometimes automatically.
When de-peering becomes contentious
De-peering becomes a contentious issue only when rationales about connectivity clash. (For those interested: I am thinking along the lines of the concept of the “orders of worth” by French sociologists Thévenot and Boltanski.) These rationales include 1. a domestic rationale, which emphasizes personal relations, reputation and trust; 2. a market rationale where the legitimate criterion for evaluation is the price and legitimate leadership equates to purchasing power and 3. a civic rationale that demands to root decision-making about interconnection in collective interest and aims for solidarity and equality. All of these notions come to the fore in conflicts about de-peering.
Let’s look at them one by one.
Personal relationships between networkers go way beyond peering (a topic for another article), but they surely accompany most peering relationships that are perceived as important by one of the parties. Some of these extend over long time. So de-peering can be emotional. One interviewee recalls a de-peering notice, “which was a copy of the email I had sent him in my previous job, just with the names changed.” Similarly, others say: “De-peering is often quite a sensitive matter”, “it is sometimes taken personally”, “usually, the counterpart is upset”. “This is not just about the hard facts of interconnecting, it is also about wounded pride.” – And on the other side of the de-peering: “My answer was: F**k you!”
One networker even described how he chose to execute a number of de-peerings by withdrawing from an Internet exchange point all together – just so to create the impression for everybody that the de-peering was impartial. He then took selected peers over to a private interconnection at another place.
It is important to note though, that not one of the networkers I have interviewed questioned that the principles of a market economy should apply to Internet interconnection, including the liberty to strive for profits, to peer freely just as long as there is mutual interest and to engage in competition. In fact, networkers commonly contain and solve conflicts within the market rationale without it even becoming public. For instance, they will clearly shut down a connection with a network that starts peering with one’s customer networks (and thereby lowers one’s profits) or, they will likely de-peer a network that enters into direct competition.
However, several networkers experience discomfort with an absolutism of the market rationale, which as a side effect, also puts pressure on what many think of as a community. The discomfort often becomes apparent in a specific situation: when a network seeks payment for a peering relationship that had previously been settlement free. Several interviewees have found themselves in such situations, on both sides. These cases are interesting, because it becomes a philosophical question whether settlement-free peering even constitutes a market relation at all. Peering does not cost the parties anything beyond connecting their networks to an Internet exchange. But when a peering is gone, it is costly to replace. In this sense, peering is quite open to definition.
Now, depending on how it is communicated, asking for money for something that was free before can lead to a showdown between networkers. There are three ways for this situation to evolve. In the first, networkers accept to read the conflict as a power play in the market. Then, a de-peering announcement will eventually lead to re-establishing the session under commercial conditions, where one of the parties becomes a seller and the other one becomes a buyer:
“When folk walk away, the other party kind of goes: ‘Well, you know, I could chase them. But if I chased them, I would be a customer.’ So if you and I are peering and you cut the wire, if it’s a true peering relationship I would go: ‘Well, I don’t care.’ But if I really needed that peering, and you cut the wire, I’m a customer. Logically, I’m more reliant on you than you’re reliant on me. You demonstrated that by cutting the wire.”
As the news travels, the paid-peering course of events reinforces the legitimacy of a market convention for others, too. The second option is for the network that was de-peered to accept the loss of connectivity, thereby keeping the definition of roles open.
Changing the benchmark of legitimacy
The third option involves a challenge: The network that was de-peered takes the position that it is not justifiable to apply a pure market mode of evaluation to this situation. It demands re-evaluation according to another rationale that it claims to be more legitimate. This is the civic rationale. We often see it when incumbent internet service providers are involved in a conflict in which they are accused of abuse of power.
The argument goes that it is not in the collective interest to let the market solve the conflict alone and that the incumbent must treat all interested peers equally and fairly. The civic rationale translates to a formalisation of rules and eventually to public intervention. In 2012, a Swiss Internet service provider was invited to change his settlement free peering connection with the incumbent into a paid peering for traffic levels that exceeded a ratio of 1:2 (i.e., when it would send more traffic into the incumbent’s network). The operator took the conflict to the Federal Competition Commission, arguing among other things that ratio-based price discrimination is at odds with the recent evolution of the Internet towards caching content at the edges of the network. The case was intermittently suspended when the incumbent became subject of a related preliminary anti-trust investigation with a third network, but has just recently been resumed.
What this example shows though, is that peering practices can only be challenged fundamentally by making conflicts known and take them to a higher level in the public sphere. Publicity is what forces the parties to justify their actions and make them accessible to a broader evaluation. However, even the most insightful networkers warn against side-effects of “well-meant” interconnection regulation. That is why calling upon the regulator more often remains a threat than that it is actually done.
Congestion is the new de-peering
With this in mind, two more recent developments may not come as a surprise:
- Peering policies are on the rise. Especially sought after Internet service providers document the criteria that another network has to fulfil to qualify for peering. This serves them in two ways: Vis à vis the regulator, the peering policies allow them to argue that they will treat every peering partner equal. And vis à vis potential peers, peering policies signal an impersonal decision process. In the meantime, every peering policy probably has a clause that allows the network to make business decisions, i.e., exceptions.
- Congestion is the new de-peering. Representatives from the largest networks were probably the first ones to realise that “de-peering between Tier 1 networks is practically impossible”. It would create a fragmented Internet and a public uproar would be guaranteed. But even with other legacy peerings “it is easier to let things live on forever than create a fuss about it or potentially have a fuss created about it,” so “we do not de-peer,” says this Tier 1 networker. Or, as another interviewee put it provocatively:
“If I am a carrier, then it is better for me to not make any statement or decision that looks like an action. […] Because if I make an action that looks like I am forcing an upgrade or forcing a network to pay, then that is potentially an area that particularly socially is kind of negative, and potentially from a regulatory viewpoint. […] So it is better for me to appear incompetent rather than actively seeking to extract a rent. So if I do not upgrade, if I have very little contact with a company, all I have to do is wait.”
Learning from bad experiences with regard to their reputation in the community, networkers who see the need to de-peer for commercial reasons may chose a softer tone to avoid open conflicts. This is actually in line with what most networkers wish for anyway in de-peering: personal communication, empathy, timely notice, appropriate grace periods that allow the de-peered networkers to transfer their traffic gracefully. What they also wish for is honesty and clarity, though, and no de-peering in disguise. In light of both regulatory threats and the fact that it is the uncertainties where market opportunities reside, the latter may be wishful thinking for the time being.
|Dieser Beitrag spiegelt die Meinung des Autors und weder notwendigerweise noch ausschließlich die Meinung des Institutes wider. Für mehr Informationen zu den Inhalten dieser Beiträge und den assoziierten Forschungsprojekten kontaktieren Sie bitte firstname.lastname@example.org.|
This post represents the view of the author and does not necessarily represent the view of the institute itself. For more information about the topics of these articles and associated research projects, please contact email@example.com.
Sign up for HIIG's Monthly Digest
and receive our latest blog articles.
We approach the de-mystification of this claim by looking at concrete examples of how AI (re)produces inequalities and connect those to several aspects which help to illustrate socio-technical entanglements.
“System Risk Indication” (SyRI) deployed by the dutch government for automatically detecting social benefit fraud. The program was shut down due to a severe lack in transparency and unproportional collection...
AI won’t kill us in the form of a time-travelling humanoid robot with an Austrian accent. But: AI is used in various military applications – supporting new concepts of command…