Tom Scholl has put up a hilarious parody of a peering cruise.
It’s meant to poke fun at the real peering cruise sponsored by Terremark, Equinix, Switch and Data, AMS-IX and DE-CIX. All of these companies operate infrastructure used by major networks to locate equipment and exchange traffic. Also kicking in sponsorship dollars is Force 10 who makes switches used by many of these organizations to connect up their networking customers.
The real event is almost as surprising (and potentially amusing) as the fake. The principle is simple: get people responsible for peering of major (and minor) networks together on a cruise ship for 5 days, all expenses paid by companies that benefit from increased peering infrastructure. Add alcohol, gambling and beach time and hope for increased peering (hence increased business for the organizations sponsoring the event). Is the decision to peer or not still driven so thoroughly by social interaction? Does the growth and health of the Internet depend upon events such as these? Isn’t it dangerous to allow network geeks anywhere near the sun?
There was a nice post on the NANOG mailing list about peering in Egypt. The poster observed “[a]s a lot of other developing countries, peering is established on personal relationships, and even with that it doesn’t work well, because both parties couldn’t agree on routing policies or even personal disliking(pathetic).” To that, the inimitable Randy Bush replied: “luckily, this kind of thing never happens in north america </sarcasm>”.
In the interests of full disclosure, I went on the last peering cruise that Terremark and LINX sponsored and was invited to this one. (I would probably have gone were it not for a very busy travel schedule this spring, a very cute baby at home and a fairly strong dislike of cruise ships). I found the trip to be worthwhile and, in the long run, useful from a business perspective. I think several other attendees I talked to agreed. But that’s exactly what puzzles me.
It seems that peering decisionmaking should have evolved well beyond something requiring large amounts of social interaction. The decision by two networks to exchange traffic should be a business decision. It should be based upon the customer base, traffic size and character, routing policies and economic relationship between a pair of networks. Ideally, it should be governed based on published policies and conducted in a relatively open manner so as to reduce confusion and claims of bias. And that’s pretty much how a fair amount of peering works. But a lot of it is still pretty haphazard and, since human beings are involved, is still based on the vagaries of human relationships.
When I wrote about the Cogent/Level(3) depeering incident, I wrote about some of the criticism of peering at the time, including: ” In our opinion, peering agreements are handshake deals that stitch the Internet together, subject to the whim of the parties to each agreement. A broken peering agreement severs the Internet, threatening the foundation of the Internet economy. These handshake deals show why lawyers are an important part of business – it’s their job to plan ahead for problems when negotiating a deal. We believe that many peering agreements were not written by lawyers.”
I was relatively skeptical about that comment when I read it. Most of the networks that I know and that I interact with are very professional about how they do peering. They have standards, policies, procedures. Many of them do not have peering agreement, but they do have a set of standards and practices that they abide by. But thinking about this cruise and the need for it makes me reflect a little bit about how much those standards reflect reality, and how much reality is bent to fit those standards.
The Internet is still very much growing up. The last 12 years or so, since the NSF began allowing the commercialization of the Internet in 1994, have been interesting and a little painful. Large-scale change is always like that. The evolution of peering relationships from something that you do over beer at a conference (or on a cruise) to something that your companies’ lawyers and C-level execs negotiate is an example of that. But that transition is mostly complete, from where I sit. Very few serious networks decide upon peering based on positive experiences at the bar, on the beach or at the pool. You explain why peering with you benefits their network and their business, and why it fits their peering policy. If you’re convincing and you have good evidence, you get your peering trial. If not, well, you complain loudly in public about how unfair Sprint is. But you still buy transit from them (or someone their size). And they know you will.
So is peering something that’s all grown up or still full of beer-driven decisions? Is the cruise merely a helpful way to entice a lot of the right people to get to the same place at the same time, or is it a boondoggle? If the process of peering is grown-up and professional, why is it even necessary to have personal relationships (forged over a margarita or three) with your counterparts at other networks? Maybe I’ll reflect on these issues on a lovely yacht cruising in the springtime just off the coast of Gary, IN. That sounds nice.