Q:

DrPeering -

Thanks for the articles - What is your definition of paid peering?

If I get free peering with ISP A, but as part of the deal they get free or heavily discounted other things from me (colocation, transit, transport capacity), does this count as paid peering?

Tuco Benedicto Pacifico Juan Maria Ramirez

Boca Raton Networks

A:

To answer this, let’s start with the definition of “Peering,” excerpts from The Internet Peering Playbook, and then talk through Paid Peering and some examples:

What is “Peering”?

Definition: Internet Peering is the business relationship whereby two companies reciprocally provide access to each other’s customers.

Internet Peering is typically settlement-free, meaning that neither party pays the other for access to each other’s customers, reflective of the underlying notion that peering is a relationship of approximately equal value to each party. Since both parties benefit about the same from the relationship, there is no need to bother with the overhead of measurement and settlement.

There is also no standard way to calculate and monitor the absolute value derived from a peering relationship. Is the value of the peering relationship proportional to the volume of traffic freely peered bi-directionally? Or is it proportional to the desirability or uniqueness of the routes? Or is the value the number of people reached?

For these and other reasons, the dominant form of peering is settlement-free. When you see the term “peering” from this point on, it means settlement-free peering.

To illustrate Internet Peering, consider the mini-Internet Peering Ecosystem shown in the figure below with only three ISPs: WestNet, MidNet, and EastNet.

Using graphical peering notation, we see that:

•WestNet is an ISP with customers shown as clouds below it, MidNet is an ISP with its own downstream customers, and EastNet is an ISP with its own downstream customers.

•WestNet is in an Internet Peering relationship with MidNet in which WestNet learns how to reach MidNet’s customers, and MidNet reciprocally learns how to reach WestNet’s green customers.

•EastNet is in an Internet Peering relationship with MidNet in which EastNet learns how to reach MidNet’s customers, and MidNet reciprocally learns how to reach EastNet’s customers.

After these two peering sessions are established, the routing tables are in place (as graphically shown as colored circles in the “routing table” beneath the ISP clouds). This diagram shows that MidNet peers with both EastNet and WestNet, and therefore MidNet customers can reach both EastNet and WestNet customers.

For convenience, we will use the simplified and equivalent notation shown in the figure below to indicate a settlement-free peering relationship between two parties.

What is “Paid Peering”?

Some companies offer a derivative service called Paid Peering.

Definition: A Paid Peering relationship is a peering relationship with an exchange of compensation from one party to the other (see the figure below).

You can see from the diagram of peering and paid peering that the two are nearly identical except for the ‘$’ flowing to one side of the business relationship.

The compensation may take the metered form of $/Mbps. In other cases, one side might cover more of the peering costs than the other.

When does peering become Paid Peering?

My litmus test: If the peering is not a settlement-free and no-strings-attached peering relationship, then it is Paid Peering.

FAQ: What is the “value” of Peering?

This question came in as well for this monthly blogger, a common thread in the Ask.DrPeering Mail sack:

“So when one party or the other determines that peering is not fair, that the peering relationship is not providing equal value to both parties, how do they determine ‘value’?”

That's a really good question. There is no single correct answer because network operators seem to determine the value of a network in different ways:

Scope

For example, a large global network operator like Deutsche Telekom has a single network (AS) covering the entire world, so peering with them provides direct access to all of Deutche Telekom customers, who live all over the world. There is significant value to a peering relationship with Deutsche Telekom.

As a comparison, other network operators split up their networks into geographic regions. Qwest, now Century Link, provides multiple aliases, one for each region in Asia, North America, and Europe. So peering with Century Link in Asia with not quite as valuable as pairing with Century Link in Europe or peering with them in North America. So the routes peered should play a central part of the value calculation.

End-Users and Capacity

The Comcast network interconnects in many regions across the United States, providing its peers with direct access to all of its customers. These customers are not all over the world but in certain regions of the U.S. The point is, peering with Comcast does not give you global reach. Comcast on the other hand has significant value as a peer, derived from the size of their customer base multiplied by the size of the pipe they provide.

As a side note, the Comcast peering policy requires that says (http://www.comcast.com/peering):

1. •Comcast requires that Applicants seeking SFI in the United States agree to provide reciprocal SFI arrangement with Comcast in the Applicant’s home market.
   

While “markets” are vague, this suggests that Comcast may be tipping their hand to international ambitions. For example, a Tier-1 ISP in Germany that peers with Comcast in the U.S., may risk having to peer with Comcast in Germany if Comcast shows up there. Interesting to see that clause in their peering policy anyway.

If you are a content provider like Netflix, and you have a lot of traffic to deliver to Comcast eyeballs, peering with Comcast provides significant value because of the lower latency and direct access to the customer base. Netflix’s recent paid peering arrangement with Comcast will provide Netlfix users with a better and more consistent end-user experience. Better for customers, better for Netflix, and one could argue better for Comcast and it’s relationship with its customers.

The Internet Peering Playbook describes the Peering Ecosystem as a collection of interdependent players with power positions, and predictable behaviors. The example I use as an example of a power position is the earlier Comcast-Level 3-Netflix dynamic. These dynamics and disputes provide some visibility into the strength of the power positions of the various players. Understand the motivations and you can predict the likely reactions and responses for any player in the ecosystem when stimuli has been placed upon it. But I digress.

Alternative Routes

The last piece of the value calculation that immediately comes to mind is the uniqueness of the route. I can only get to France Telecom customers through France Telecom. That increases the value of peering with FT since there are only worse paths to those customer eyeballs. You either peer, or there are hops. Compare that to the value of peering with a Tier 2 ISP that has 70% of their customers multi-homed to another upstream provider. It is better to peer, but if you won’t peer with me, maybe your competitor will.

Traffic Volume

Another variable of course is the volume of traffic that will be peered away, as this affects the money saved and how bad it would be if the peering went away.

Peering Ratios

Yes, some network operators use traffic ratios as a proxy for value symmetry, and there are some logic problems with this, but I’ll defer that discussion to another blog.

FAQ: Who sells Paid Peering?

FAQ: Do all eyeball networks offer paid peering?

No, not everyone markets and sells Paid Peering. I asked some of the ISPs why they don't have a paid peering service. The answers were really interesting:

BellSouth (before they became AT&T) said they would not consider offering a paid peering product? The network engineers over lunch shared that it is hard enough to train salespeople to sell the differentiating benefits of their transit service. If they offered a paid peering service, would be more difficult to sell, would compete with their transit service, and most likely would be ignored by the market or cannibalize their transit business. For these reasons, they choose not to offer to sell Paid Peering.

The BellSouth people pointed out though, that anyone can construct paid peering from transit merely by filtering out all but customer routes. And effectively this would provide paid peering without any of the downsides described above.

FAQ: Who is most suited to buying paid peering?

I am working with a number of gaming companies, financial institutions, and social networking companies that will benefit from buying paid peering as part of their blend of Transit, Peering, CDN, Remote Peering.

The gaming company application, for example, requires low latency, and more importantly, low packet loss and low jitter. These messages after all may determine if their gaming character lives or dies! So the gamer and gaming company have a vested interest in the best possible performance. In the case of financial trading, there is real money involved in the speed of the transactions. They shave a millisecond off a network connection and it could mean millions of dollars. They don’t care about the costs; it is the end-user connectivity experience. And social networking sites in general are very adept at tuning their pages - in some cases they employ massive clusters of computing systems in place to construct their pages on-the-fly. But, they are usually very efficient encodings, and so are small pages that can be delivered really fast.

Peering, paid or otherwise, gets the content as close as possible to the eyeballs, and this is worth paying for in several applications examples.

Second, since their encodings lead to small web pages, their traffic volume is pretty low. This makes it difficult to get the interest of some of the eyeball networks. Even the more open of the selective ISPs might say, “You know, call me back when you can peer with me 1 Gbps.”

Third, if the price of paid peering is around the price of transit, and it gets you better performance for about the same price, so why wouldn’t one buy paid peering?

Finally, again because of the small efficient encodings of the web objects, the traffic volume is small, so the monthly cost of paid peering is low. So low in fact that they might not even care if the price is a little higher than the cost of transit - when multiplied by the minuscule volume, the monthly cost of paid peering is pretty small in the scheme of things.

FAQ: If Comcast started this paid peering thing; will the other cable companies follow?

Well, first off, Comcast did not invent paid peering. A variant of paid peering has been around for a long time. I remember researching paid peering and learning that AT&T had a paid peering service, and that AT&T and the paid peers viewed this as a stepping stone towards free peering. (Some will contest this positioning but...) The problem I had was that I couldn’t find anyone willing to talk about their paid peering relationships.

Paid Peering is like paying for a date to the prom – neither side is inclined to talk about it.

Neither side has incentives (or are contractually allowed to talk about) the terms of the deal. And this makes it very difficult to understand to describe, to quantify and discuss these specific cases of paid peering.

FAQ: Is Peering Barter?

Peering sounds like barter, something we tax in Australia. Is peering barter?

I received this question in Australia and I believe there are at least a few differences between peering and barter.

First, barter relates to goods that are commodities, indistinguishable from one another. Some would say that each network brings different scale, scope, and qualities to the table that make peering with some better than peering with others.

Second, if one was to tax barter, one could take the market value of the commodity and tax as if the commodity was purchased on the open market. But what would the price be for taxing peering? The cost of the next best alternative, Internet Transit? There are a vareity of things (commit, locality, traffic type eyeball vs. content, etc.) that would affect the price of transit. Some ISPs, in order to improve their traffic ratios, would provide free transit to eyeball heavy networks for example. So would the tax price be $0?

Then there are the many ambiguities surrounding measurement and audit. For example, there is all of the ambiguity about which networks from which countries would have to pay the barter peering tax, etc. and since companies view peering as a free and reciprocal arrangement, they don’t robustly measure it for billing purposes. This is a rat hole.

In general, these things are so fluid now, any barter tax system will be gamed by the network operators affected.