Todd Underwood posted a blog entry on the Renesys Blog entitled Cogent Becomes Transit-Free. It’s an enjoyable read for those of us sick enough to be fascinated by BGP and peering. In it Todd elaborates on Cogent’s acquisition of PSINet, their use of transit providers, and numerous peering disputes for which they’re infamous. Fun for the whole family!
Cogent’s peering status is especially interesting in light of the sort of conclusions that CAIDA’s 2007 AS Core Visualization project reached. It showed Cogent as one of the top networks, up there with the likes of Level(3), AT&T, Verizon, and Sprint; they were ranked well above many transit-free networks. At the time I thought this was notable given that Cogent wasn’t transit-free. Which, as the devil’s advocate, led me to wonder as I read Todd’s article about whether this latest milestone is really that meaningful.
Becoming transit-free is certainly important to Cogent because it represents the reduction/removal of transit services costs. But with today’s low price of transit bandwidth often approaching the underlying cost of providing the service, this has got to be a very marginal improvement to their P&L. Clearly this is Cogent’s business: pushing down costs in order to be the low-cost Internet provider. But for the rest of the Net, does transit-free matter?
I suspect that Cogent’s service today performs only slightly better than it did before it became transit-free. And that’s possibly not even true depending on how the routing change affected their backbone’s traffic load, how fast their routers and transmission paths are compared to their former transit provider(s), etc. For an existing well-connected network, the marginal improvement gained by direct peering versus transit through an equally capable network partner just might not be worth the effort unless there is a large volume of traffic involved. I’d be surprised if this was the case with Cogent-ATDN traffic.
So what is the value of being transit-free? In today’s consolidated industry perhaps the biggest networks carry enough traffic to care about the marginal cost improvement. But at what size does that begin to matter, and what does that suggest for tier-2 networks that peer extensively? I don’t know the answers, but I do know that the assumptions which were true 5 years ago need to be reconsidered.
2 responses so far ↓
1 bensons // Jul 3, 2008 at 19:45
I just came across a relevant blog post at http://www.telecomramblings.com/2008/07/carriers-and-replacement-value/.
To quote:
“Some go further with this contention, that the high replacement value is an *advantage*, a moat for the competition to cross. Well, there is a silver lining for everything I suppose, but what needs to be acknowledged is that this moat exists solely because of the failure of the assets to justify the investment. OF COURSE nobody will enter a market that isn’t strong enough to generate financial returns on an investment. If it were a good market, then these companies would have valuations above replacement cost, and suddenly people might think about entering alongside them. Having a moat because your sector is unattractive isn’t really where one wants to be.”
2 todd // Jul 9, 2008 at 7:38
benson:
thanks for referring to my blog.
you may want to be careful about the distinction between transit-free and settlement-free. cogent is almost certainly not settlement-free. they probably have a customer relationship with sprint at the very least, and quite likely some others. i may have been insufficiently forceful about making that distinction in my blog posting.
the rest of your comments apply to many other networks, though. It’s not clear what the market or business value of being settlement-free (or transit-free is).
the one clear case i can think of: being transit-free gives a network stronger negotiating position wrt peering disputes, since a negotiation partner cannot force traffic down an unwanted path during a depeering.
You must log in to post a comment.