SFI/SBI/Transit - Dumping

Hello all!

I’m going a bit deeper into the study of Peering Relationships…
And one of the possibilities that I’m trying to understand better on the Peering Relationships on the Internet been considered dumping(economy).

The matter here is more on the economic and commercial aspects than on the technical side.

On the scope of Internet Peering Interconnections(Transit, Settlement Free, Settlement Based.):

  • There are any public papers, or even judicial processes, about situations that the possibility of Dumping(economic) was supposed? Independently if de dumping argument was accepted or not.
  • Does any colleague could remember some memorable case about that?
  • How does it affect the Big-Traffic-Guys (OTTs, Hosting/Colo/Cloud Datacenters, CDNs) and their routing policy on PNI/MPLA, SFI/SBI?

Can you explain to us amoeba-like salesmen what is dumping? I assume it involves highly unbalanced traffic flows.

-R.

Hi Douglas,

I'm not aware of any examples of Dumping (in the economics sense)
involving backbone BGP peering. Usually the problem is collusive
exclusion.

If you want to cast a wider net, Erols Internet of the 1990s offers an
interesting case study in driving competition out with extended
below-cost pricing. But this was dialup and DSL service, not backbone
peering.

Regards,
Bill Herrin

I was an Erols customer during that time. What’s the story?

There is no specific story on the focus.
My objective is to go a bit beyond the technical aspects of Peering, or De-peering.

At the first moment, I don’t mention some cases that I have in mind, exactly to avoid the polemic and focus on the aspects around the cases.

I will give a hypothetic example, with no real names:

* fischerdouglas@gmail.com (Douglas Fischer) [Tue 16 Mar 2021, 11:32 CET]:

And then??
Can this be considered an anti-competitive act?

I think you're asking this on the wrong list. We're network operators, not lawyers with a specialisation in competitive markets regulation.

  -- Niels.

Well, yes and no...

No, we aren't lawyers, but yes in that there is some merit to the question from a network operator standpoint.

Sure, the CDN operator wants to deploy their node with as little cost to them as possible, particularly if the target country does not move the traffic and $$ needles for the CDN operator, i.e., they want to de-risk their investment as much as possible, hopefully until traffic and revenue picks up from that specific market.

Getting a local ISP to shoulder that responsibility, likely disguised as "content is king, so you benefit" would not be unusual.

It all comes down to how desperate the local market is for the content, and how far they are willing to negotiate in or against their favour.

Mark.

Hi Douglas,

The words you're using here are very strange but I think you may be
trying to ask if "settlement free peering" can be anticompetitive.
Yes, it can, but from the way you talk about it I think your
understanding of when and how is probably upside down. It has little
or nothing to do with dumping or other forms of cross-subsidy.

Let's start with some terminology.

"Settlement free peering" is business practice in which two
organizations place routers physically next to each other* and trade
network packets which are sourced from one of the organization's users
and destined to the other organization's users. This differs from a
normal business relationship between Internet companies in several
important respects.

1. Both organizations pay their own direct costs to place their
routers next to each other.

2. Neither organization pays the other any money. They do not "settle
up" with each other depending on the packets transmitted, hence the
packet trade is free of any settlement settlement process.

3. Only packets which are from one organization and to the other are
transmitted via the connection. Unlike a normal "transit"
relationship, the settlement free peering connection links neither
organization to the Internet at large. Rather it connects them only to
each other.

The idea is that directly connecting both reduces cost for both
parties and improves the customer's experience by providing faster
data transmission between the parties.

Organizations which engage in settlement free peering generally have a
"peering policy." This policy describes the circumstances in which
they seek to or are willing to establish settlement free peering with
other organizations. Their peering policy is said to be "open" or
"closed" depending on the manner in which they select their peers.

An "open peering policy" is one in which, within reason, an
organization agrees to engage in settlement free peering with every
other organization which asks to do so. The two organizations either
place routers in a mutually agreeable location or the asking party
places a router somewhere that the asked party is already present.

A "closed peering policy" is one in which the organization is very
selective about who they will establish settlement free peering with.
Even if the other organization is willing to place a router next to
theirs, they will only trade packets for free if the other
organization has specific business characteristics which meet with
their approval.

Finally, "paid peering" is exactly like settlement free peering except
that one organization pays the other for the connection.

As you can probably guess from the terminology, it's practically
impossible for an organization with an open peering policy to drive
anticompetitive behavior with settlement free peering. The same
business relationship is open to all seekers; they have but to ask.
The question gets more interesting where one or both have a closed
peering policy and offer paid peering to organizations who do not
qualify for settlement free peering.

Now, moving on from fact to opinion, closed peering policies can be
and often are anticompetitive. The peering policies typically have two
components: high data rates (exclude anyone who isn't already an
industry titan) and comparable bytes in and out (make content
providers who send more than they receive, like Netflix, pay us).
While industry leaders don't negotiate the list of folks they will and
won't peer with, they nevertheless all pretty much arrive at the same
list.

Anyway, hopefully this information is helpful to you.

* The routers don't necessarily have to be physically right next to
each other but that case is easiest to understand and the variants
where they aren't are functionally identical.

On a related note, it's been my experience that while cross-subsidy
exists in the Internet transit industry, it's not one of the critical
sources of anticompetitive behavior. The two primary sources are
things like the subtle collusion involved in closed peering policies
and product tying where many valuable services like a wavelength on a
last-mile PONS line cannot be purchased independently of the Internet
service lighting that line.

Regards,
Bill Herrin

There is no specific story on the focus.
My objective is to go a bit beyond the technical aspects of Peering, or De-peering.

At the first moment, I don’t mention some cases that I have in mind, exactly to avoid the polemic and focus on the aspects around the cases.

I will give a hypothetic example, with no real names:


The “SteveAndEdNet”, a CDN Company, decides to extend its branchs until “Kingdom of Far Far Away”, and creates a POP there.
Installs itself on “DorisInnDatacenter”, connects with some IXPs over there, connects some PNIs with some nobles.
But, after some time, “SteveAndEdNet” make a deal with “RumpelstiltkinNet” a Transit provider that operates there.

On that deal “RumpelstiltkinNet” ensure to supply the traffic demanded to “SteveAndEdNet” POP to be considered technical justifiable…
But it also demands, that “SteveAndEdNet” drain all the traffic to IXPs and PNIs…
With that, “RumpelstiltkinNet” can be the only one reseller of the content delivered by “SteveAndEdNet”.

This is a foggy example that I’m trying to understand better by the point of view of Economics Dumping.

And then??
Can this be considered an anti-competitive act?

If anyone feels more comfortable reaching me privately, no issues with that.

Hi Douglas,

I’ve been involved in negotiations with country-level providers that engaged in negotiation tactics like that.

From the content-provider side, such demands (ie, drain all traffic away from IXPs and other SFI PNIs)
might last a short time, if RSN (RumpleStiltskinNet) undercuts the pricing so low on their connectivity
that it makes sense to move all the traffic over, or bundles in other products (power, space, remote
management, etc.) so that the overall package price is lower than it would be if the IXP and SFI PNIs
were maintained.

However, such relationships tend to fall apart the first time RSN has a large-scale outage
that negatively impacts the CDN and all its content customers, at which point the
negotiation table is usually re-opened, and the CDN says “in order to protect our
customers from your ineptitude, we require alternate pathways for the content to
be served in your region” – and the exclusive arrangement goes away.

The simple fact of the matter is that in general, agreeing to single-home your
content behind a single provider results in reduced availability for the content, as
a single provider has less redundancy and less diversity than a blend of multiple
providers, whether paid or settlement free. And once a major outage happens
that threatens the top-line revenue numbers for the CDN, saving money on the
bottom line suddenly becomes much less important than preserving the top-line
revenue number.

There’s a finite and limited benefit RSN can offer by reducing cost numbers;
but there’s a potentially unlimited upside risk. That is, RSN can’t reduce
your cost numbers from what they are today to less than zero; so there’s a
maximum benefit they can offer. But if they have a bad week, and you go
dark, and lose all your customers, you could lose all your revenue–and there’s
no limit on how much you could lose on that end, as your customer base and
revenue numbers go up.

Thus, on balance, such agreements don’t last, as either the CDN goes out
of business, and the agreement thus ends, or the CDN becomes more
successful, and the top-line revenue dollar value risk outweighs the
offered cost-reduction saving benefit RSN is bringing to the table,
and the agreement is thus terminated.

I hope this helps clear up the situation for you. :slight_smile:

Thanks!

Matt