Ask DrPeering

Dr Peering -

What is “Paid Peering” and when does it make sense? Why won’t anyone talk about it?

Gavdi Enron

------------------

Mr. Enron - Good questions.

DrPeering has been searching far and wide for evidence of Paid Peering, from the well connected peering ecosystems in the U.S. and Europe, all the way into the deepest darkest parts of remote peering ecosystems in south east asia. There have always been rumblings, but no one was willing to talk about it and be cited.

Let’s first remember that Internet Peering is defined as a reciprocal exchange of access to each others’ customers. This is typically a free relationship, as both parties feel they are getting approximately equal value in the deal.

Next, recall that Internet Transit is defined as a business arrangement whereby on Internet Service Provider (ISP) provides (usually sells) access to the global Internet. Here there is a clear customer-provider relationship; the customer pays who can get traffic onto the Internet.

These working definitions were more or less easy to observe, document, verify, and validate with the Peering Community. But Paid Peering has remained elusive; veiled behind NDAs and the closed doors of peering discussions. Over the last ten years Dr. Peering has spied some paid peering patterns of behavior in the ecosystem.

Paid Peering Discussions

Here is the typical sequence of events that lead to Paid Peering Discussion.

Peering is requested. To get free Internet Peering, two ISPs explore the mutual benefits of the proposed peering relationship. Network maps and mtrg traffic graphs are exchanged. If one or the other party feel there is not mutual benefit, if they are not indeed ‘peers’ in the value they bring to the relationship, and the peering doesn’t typically happen.

So, Peering is denied. Now, in some cases, the party denying the peering request suggests that the denied party buy transit. In some cases, the party denying the peering request suggests that the denied party buy “Paid Peering”, which is exactly like Internet Peering except that some money is exchanged.

In both of these cases, the denied party saw themselves as a peer, but were for some reason rejected.

When peering is denied, it is not surprising that

1)Neither party wishes to talk about the denied peering event. (No one wants to be seen as a network not worthy of peering, nor be seen as the “bad guy”.) There are the occasional whiners who complain publicly in the hope of getting sympathy, but Dr. Peering has never seen this leading to the peering that was originally sought.
2)The denied party feels slighted, and maybe even the employees of the company feel somewhat inferior. There may be anger, embarrassment, disappointment, resentment, tears in beers.
3)The denied party sometimes responds to the suggestion to “buy transit” or “buy paid peering” with one half of the victory salute. (For those who never heard this expression, it is an obscene gesture.)

Sounds like dating, no? Dr. Peering thinks so.

But Paid Peering may be emerging now as a product that makes sense for a particular segment... but more on that below.

Let’s speak to your next question...

Why don’t people talk about paid peering?

Those who sell paid peering will talk about it, but

1)You have to find the right people...They often don’t have a dedicated commissioned sales team for the product. Sometimes, this is because they offer the product as a stepping stone to free peering, once the prerequisites are met.
2)They get higher margins by selling transit, a well known product within the company. The margins and therefore the commission for the salesperson who bothers to learn about the product may be small, and the complexity of the sale may be greater.
3)They offer it at a price higher than the market price of transit. ATDN was famous in the Peering Community for this. The resulting “Are you crazy offering it at that price” doesn’t lead to many sales, so there is a disincentive to bringing it up. The counter argument is that the quality of the end-user experience is better with peering, so those that care about that will pay.

For these reasons, we haven’t seen a lot of force behind the paid peering initiatives in the peering ecosystem.

It is worth pointing out too that those who buy it generally don’t talk about it because by doing so, they are admitting that they are not peers with their paid peering partner. You sometimes see Non-Disclosure Agreements in place with Paid Peering Agreements as a result.

Why don’t all ISPs offer a paid peering product?

Dr. Peering asked many telcos about this and the answers were pretty consistent.

The most common answer was that it was too much effort for the telco to launch a new product, especially when it competes with a product they know and sell today (Internet transit).

The second most common reason not to offer paid peering was that there would be too much cost to offer the new service, relative to the margins they should be able to get.

Consider all the internal new systems to sell, monitor and debug, train for, pay commissions on, report on, etc. All of this cost, for a service that the market believes should be priced at less than the already low margin transit service.

These companies told paid peering prospects to buy transit and filter out the routes that are not wanted, as illustrated to the right.

When does Paid Peering make the most sense?

The best case for paid peering is starting to appear today: high-speed, high capacity last mile access.

Comcast for example has a paid peering business arrangement where companies can directly connect their content or CDN services into the Comcast last miles, and those doing so reported to DrPeering that they are getting 3 times better performance as a result. The better performance is compared against sending traffic through a potentially circuitous path through potential many ISPs that are managing their traffic to maintain their peering ratios.

Paid Peering with last mile may be the most effective weapon for Content Delivery Networks. Rumors are that the price of paid peering with Comcast is very competitive when compared against transit. DrPeering suspects Comcast will succeed in helping this industry uncover and embrace the Elusive Paid Peering.

End Notes.

There is actually another option seen in the wild along the continuum of alternatives between peering and transit: partial transit. Partial falls into three categories - exchange X only, region X, all peers

The first example of Partial Transit is when an ISP sells access to the routes it learns at an Internet Exchange. They can say “Hey, you are thinking of joining the GINX, but you will have to pay the membership fees, the colo costs, etc. and then establish all the peering relationships. Instead, I will deliver to your router all the routes I learn from my long lived presence there, all for less than you participating there yourself.”

The other two partial transit examples are similar but for the regional routes, and for peer routes - only routes learned from the peer’s peers.

James Blessing (consultant, Garou Ltd) points out “networks offering partial transit can have problems with their peering arrangements if they don't make sure they are in line with peering arrangements that include 'universal announcement' clauses."

He goes on to say “Another reason for paid peering is that as you are a contractual

partner of the company (i.e. have a contract with monetary values) its easier to get

capacity issues and network problems (such as ddos mitigation) addressed quickly.”

“For filtering of routes from a transit feed you really need good quality (documented

and maintained) community support within the transit supplier for this to work - is there

an article on the benefits of communities for the peering community somewhere?”

“As a well designed and maintained community structure can make sure that path

selection can take geographic impacts into account - i.e. not having traffic between

two uk networks travel to de-cix to be exchanged.”

Will Hargrave points out that the AS-PATH will be longer in the partial transit solution than the direct peering interconnect, and that it will be more difficult to migrate to a more fully peered network.

Pricing Paid Peering. The metered price of paid peering and partial transit has generated a lot of heat. Many believe paid peering is a subset of full Internet transit, and therefore should be priced a a fraction of the price of Internet transit. Others believe that peering provides greater performance and therefore ought to be priced higher than Internet transit.

In 2009, in the market where Internet transit can be had for $6-$9/Mbps, paid peering seems to be pricing out (and catching on) at $1-$3/Mbps.

Some Notable Paid Peering Providers

AOL http://www.atdn.net/paid_peering.shtml

Comcast http://www.t1r.com/client/view.php?rid=56457

Verizon http://blog.streamingmedia.com/the_business_of_online_vi/2009/01/verizon-cdn-announcement.html

References

Peering Coordinator Panel ; 28-Nov-2007 http://www.peering-forum.eu/epf2/presentations/7.PeeringPanel.pdf

AOL Paid Peering POP and Locations http://www.atdn.net/paid_peering_locations.shtml

Verizon Port Partner Program - Just Smoke? http://www.telecomramblings.com/2009/01/verizon-port-partner-program-just-smoke/

Net Neutrality and the Carrier Hotel - CRGWest - http://www.crgwest.com/PDFs/NetNeutralityandtheCarrierHotel.pdf

asr’s Paid Peering Blog http://www.voxel.net/blog/2009/01/paid-peering/

Complexity of Internet Interconnections: Technology, Incentives and Implications for Policy http://people.csail.mit.edu/wlehr/Lehr-Papers_files/Clark%20Lehr%20Faratin%20Complexity%20Interconnection%20TPRC%202007.pdf