I have some interesting perspective on Cogent. We use them for a couple megabits at a colo facility for hosting galleries. They're actually not bad. The problem we have is with the facility - they REALLY SUCK. We're talking timeouts within the facility and hours of downtime due to fiber cuts in the facility and tripping over the cables etc. But the Cogent has not been the problem yet and we have been there for about six months. FYI today we're missing a few hits coming through cogent but it's not that severe.
My criticisms of Cogent are things you would not normally think about. First, run a trace off their network. Their backbone looks kind of like a giant figure 8 -
http://www.cogentco.com/home.html . This is a problem. Any one point on that network has only two points of failure upstream. If one point does fail, it cuts off the entire flow in that direction and forces the rest of the path to reverse and depending on the location of the failure can force almost the entire half of the country to follow the working path. If they're operating near 50% or higher capacity on the middle states this can cause full network outage. That's the theory anyway.
Another downside to the topography is that the packets have to take an inordinately large number of hops to get to a peering point. Midwestern facilities have to run many hops to get the packets out. That's a problem as it takes time and adds many more points of failure. Plus if you look at their map it's seems that they're forcing all the traffic in the northeast to come through the same pipes. Currently Cogent is running slower for the same geographic distance than most.
Cogent has also repeatedly refused to answer questions about their peering to me. They are especially tight lipped when I ask about oceanic peering and transit. That's not a good sign. I have been talking to a couple other providers about them and they said that when they switch people from Cogent to another provider that their traffic goes up. They all attributed it to international users that were timing out from bad peering overseas.
The $1000 a month plan there is for people who do not host. It's for people who pull down more than they push up. When you're using Cogent's model, the hosting people are the first to jump on board as well as the dominanant users of bandwidth. When you try to set up a peering you have to be on the same level in that area as the people you're trying to peer with. The big providers don't want to peer with you if all you're doing is hosting. They want to sell you a line, which is pretty much what they're doing if all you're doing is pushing. So Cogent encourages apartment building owners and companies to pull so they can get their ratios in line so people will peer with them. It's pretty rational.
My vision of Cogent in the future is very good. They have a system in place that is years ahead of everyone else. They came in after everyone made their mistakes and have a system with significantly lower costs than anyone else. But more importantly is their effort they are making with the last mile, the apartment buildings and offices. If you control the last mile you control the game. Ask Walmart about that strategy. My belief is that if they don't screw up, the concept is quite sound - they will have people hosting on their network and they will have people pulling from their network. Eventually everyone will be forced to peer with them to get access to their apartments and offices, and they'll be one of the most relevant players. The sooner they get their last mile customers, the faster this will happen. It's a matter of time and not screwing up. And I would like to see them get that backbone map looking a little more appealing.
Buy my content, we know what we're doing.