Also, at Bell you can buy 40GB extra for $5 if you pay in advance, which equals $0.125/GB.
http://www.bell.ca/shopping/popups/personal/internet/usage_e...
This is how every single element of the economy works.
Plus, if the ISPs really want usage based pricing, then they should do that instead of this arbitrary download cap stuff that is nothing about usage based pricing.
Most quotes I've heard recently put the cost of bandwidth at around 1-2c/GB. I assumed, quite possibly erroneously and I'll go hunting for the source in a moment to confirm/refute said assumption, that when TekSavvy's CEO said that a conservative estimate (read: "highest reasonable estimated price") of bandwidth could be 30c/GB, I was assuming that he was taking things like administrative costs and possibly infrastructure costs into consideration. I made this assumption because of the extreme disparity between the two cost points - 1c/GB from some versus 30c/GB from him.
I realize that I didn't make this clear in my original post - I didn't even hint at it, actually - so I apologize for that.
Also, when I just now checked my wording in my original post, I noticed that you edited your comment to include a back-of-the-napkin estimate of Bell's incremental cost of delivery to be around 13c/GB. In my opinion, this only strengthens my assumption, since I sincerely doubt Bell would provide additional bandwidth for less than cost.
Looking at Bell's website right now, I see that their "Performance" plan in Ontario lists 6mbps speed with a 25GB cap for $31.95 per month.
Using your calculation of cost of bandwidth being 13c/GB, this means that full utilization of the cap would cost them $3.25 per month, leaving $28.70 for administrative costs and upgrading.
In a completely unfair comparison, I'll mention that Rogers charges me $6.95 per month (last I checked, anyway) as a usage fee/basic fee/system access fee. Let's nearly double that up to $12.72 (to make the next calculation easier) and call that the cost of doing business (tech support, paper pushing, general administration). This brings the total up to $15.97 a month, which is HALF of what Bell charges for this plan.
Bell then gets to choose how much of their 100% profit they want to use to upgrade their network, and how much they want to use to give their CEO a raise.
As far as the overage is concerned, let's note that when I called bandwidth costs 13c, that was on the assumption that Bell made no profit from their "insurance" option. Anyone who doesn't pay insurance gets to pay $2/GB, of which at least $1.87 is pure profit. Their overage would be 1538% of cost - a nice markup.
Edit: I just noticed[1] that the true cost of this plan is actually $41.95, but they offer $5 off for the first twelve months and an additional $5/mo discount to be used towards their satellite service. I'm not going to bother doing more calculations, but I thought it was worth noting.
[1] http://www.bell.ca/shopping/jsp/pageblock_styles/Pricing/fle...
(I had to disable JavaScript to get that link. The JS-Enabled version requires you to drill down two links to get the full details to appear, so it was impossible to get a direct link to the JS-Enabled version of these details)
At the wholesale level, bandwidth is sold as a committed throughput rate. Essentially like an "open pipe" of various diameters (to use a semi-bad analogy).
Wholesale bandwidth is currently around $15/Mbit/mo. This means that for $15 you can get a 1Mbs connection to a wholesale Internet provider for 1 month (not really, because they generally want you to buy at least a 10Mbs handoff). If you leave this connection completely idle, you pay $15/mo. If you saturate it 24/7, you pay $15/mo.
Because of the bursty nature of Internet traffic, ISP's will typically resell multiples of their committed bandwidth. You might have a 100Mbps handoff, but could sell 500 or 88Mbps worth of traffic to your customers.
Bandwidth really doesn't get metered down to the MB or GB until you get to the consumer level. This is mostly done to "shape" demand somewhat, and allow for a billing method to charge users who are heavy users (eg: if your customers all tried to actually use that 800Mbps you were selling them and you really only had 100Mbps of available bandwidth, you're going to have to do something about that...). For higher-level datacenter type customers, it's typically a 95th percentile billing method, not a direct $/GB overage charge.
You can do some math and take your operating costs and divide by your bandwidth and figure out your cost/GB, but that's not going to be highly accurate. In many cases you can go from say a 40Mbps connection to an 80Mbps connection with all the exact same equipment, and just pay the upstream for the extra 40Mbps/mo. This would mean the same infrastructure cost is divided over a larger amount of bandwidth, resulting in lower cost/GB (until you need to add another card to your edge router, or add another router altogether...).
I think that part of the problem people are having with coming up with fair and realistic pricing for bandwidth is that unlike gas or water or electricity, you're not really delivering a physical commodity with costs that vary directly by quantity used.
I've always known that there was something inherently wrong with charging by traffic, but I could never put my finger on what it was. A clear explanation, such as you've just provided to me, on exactly how the internet works and is billed at higher levels makes things obvious.
Traffic is a problem, but only insofar as it contributes to saturation of the overall available bandwidth of the existing infrastructure. Calling the problem "traffic" is just a convenient scapegoat for other concerns that can't be directly billed as easily, even if traffic isn't the real problem.
Thanks. I'm always impressed by the quality of discourse on Hacker News.
