Originally Posted by Phil21

This is a funny thread..

All I can say, is the $399 for 10meg price point 2 years ago, was a huge gamble. Today, it's definitely profitable IF you have a decent sized customer base.

Ask any of our customers with this plan if they can or cannot push 10meg.. They can, and we even allow uncapped usage on these, overage based on 95th percentile.

I think some people simply do not realize the scale that many hosts are at these days. Also keep in mind not all traffic goes out transit links, if you're anywhere near competent. I can't divulge our peer:transit ratios, but I think a lot of the much smaller hosts would be amazed at how much traffic you can peer off settlement free once you get to a certain point (without sacrificing performance, most of the time it's actually an improvement).

Really it comes down to hardware pricing, datacenter (space/power/cooling), and people costs. The bandwidth costs are somewhere around 10-20% of a 10mbit sale. Sure, the profit isn't amazing for each machine, but aggregated into 50+ the numbers really start to make a lot of sense.

I can't speak for other hosts, but we don't "oversell". We carefully manage growth so that no trunk ports are ever contesting for bandwidth. We also maintain more than double our *bandwidth commitments* (not actual usage, which is of course less) in internet facing transit capacity. We could have every person burst to 5 times their commit on a given aggregation switch, and we'd still have plenty of headroom. This plan allows us to grow WITH our customers, instead of disrupting their business to play musical ethernet ports.


Plus, when you get to a certain size, and have enough technical ability large carriers will cut "special deals", which I also won't get into. However, you need to have an actual engineering staff who knows wtf their doing and has a good network of colleagues at other companies for these deals to even present themselves to you. The game definitely gets interesting at a certain level.

Again, bandwidth on a 10mbit plan is a very small part of the overall cost matrix. People by far is the largest, and equipment/maintenance of said equipment is the second largest. Equipment costs can be somewhat defrayed if you buy in enough volume (say, commit to delivery of 200 servers/mo) from a large manuf. like Dell or HP and have the credit to get a major lease signed. I know hosts (not us, we don't push enough volume, since this isn't our business focus) who can lease dual xeons for less than $43/mo each...

Still think the model is unworkable? I used to say the same thing, when we were much smaller. However, as you gain experience and knowledge of how things work, the numbers definitely start to make sense.

I'm not saying we're the greatest ever, but we work hard at it, and are improving every day. The most least profitable plans for us are the tiny little commit folks (1-2mbit), since there is almost zero margin in them as bandwidth pricing scales much better than hardware/staffing..

Just some food for thought.

-Phil