How much does 'bandwidth' really cost?

Ideally someone here might actually have experience working for an ISP or telco but either way I’m lookin for info on the approximate prices of bandwidth in north america. For instance, when I lived in a university residence, the admin only let each user of the school’s oc3 (or whatever it was) use 750 megs a week. Theoretically this could have been to prevent downloading of pirated software & movies but I believe it was actually related to saving money. So yeah, does anyone have an idea how much 1 gig of bandwidth is worth, it probably varies a little with the technology being used to download (isdn, dsl, etc) but lets assume we’re talking about t1+ lines, how much is each gig worth approximately?

cheers.

maybe the question was kinda confusing? Please ask if any clarification would be helpful…i’m basically curious about bandwidth as a commodity.

Let’s start with a phone line and work up from there. There are going to be actual Telco guys and gals along to put me in my place.

A phone line is a DS0 and is 64k. You’ll get 24 of these in a T1 or DS1 (Digital Signal). Now you get 28 DS1s in 1 OC1. From here on up an OCn is n* the OC1 in terms of bandwidth. So we have 24283 = 675 phone lines on an OC3.

Now let’s say a domestic call is $0.04 a minute, so an OC3 would carry $27 worth of traffic a minute. But you want total data transferred not transfer rate. So an OC3 (~50mbps) over 60 seconds has transferred 3Gb of data. Together this gives us $9/minute per Gb of data.

>_<

Nine dollars per gigabyte? That can’t be right, can it? I guess four cents a minute is reasonable for long-distance, domestic calls, but surely the ISPs get some kind of bulk deal?

Possibly, as I said there are a few telco people on the board who may be able to give better pricing. Though, I believe you’re right about costs for bulk line rates.

Oh and it’s $9 a Gb regardless of how long it took you. I’m not sure how I managed to get /minute in there.

I do not understand Grey’s example at all, but basically, the more bandwidth you buy, the cheaper rate you get it for. Sellers have plans - this kinda connection, with this much throughput (might be measured in % or real Gb), this much per Gb / Mb.

This is the same as anything - you can buy a 10kg bag of gravel from your corner hardware shop for $10; buy a ton from a garden supply company, and it’s not going to cost $1000! (it’ll be cheaper!)

I use about 1.4Tb (terabytes) per month with my website, and I pay around $1US per Gb. That’s a pretty normal fee for a big web hosting company. Built into that is a level of service (for example, they have a 24/7 service centre, if my site has a problem, I can call them for help).

So, bandwidth you buy has different qualities. You can get cheap bandwidth that might be of low quality (slow, high loss, intermittant, whatever) (heh, actually, you can get expensive bandwith with the same problems, sometimes!).

Being a university, they probably got a good quality feed. They’d have a fair idea of how much each faculty used last year, plus how much the residences used, so when organising a new plan for the next year, they could get a good deal from the ISP they use. Uni’s’d use an ISP the same way anyone does, from you, a small business, big business, and government departments (with exceptions for some of the bigger uni’s who co-developed the internet with whassname).

abby

A little confusing I admit. Basically the $0.04 price point for a phone call was all I had. Then I just extrapolated up to the OC3 rate that was presented.

Since you’re an actual paying consumer of data your numbers are more reliable. I’m just happy I’m in the ball park.

We don’t charge per bit or byte. We apply two charges to our customers. The first is a per month charge for the physical line. A T1 goes for something like $600.00 and for that you get a best effort QoS (quality of service). It’s higher for better QoS. What QoS you need depends on what kind of traffic you have. If you are running delay sensitive data like VoIP telephony or video conferencing, you’ll want as little delay as possible so you’d probably want to pay for best QoS. If it’s just internet, email or cash machine terminals, then our “best effort” service would serve just fine.

The 2nd charge applies if you are a host that wants to serve multiple remote locations. You’ll need to divide that T1 up into channels and allocate each remote location a certain amount of bandwidth on the T1. This is a small charge, just a few extra dollars per month I believe.

But I can tell you for sure that we don’t charge by the byte. I can see your bits & bytes moving through the public data switches, and I can see when we start policing if you try to exceed the contract you’ve paid for, but that’s it.

Lots of technical details omitted because they are boring. :slight_smile:

Bandwidth is essentially free, its the pipes that cost money. The largest cost for data is laying cables everywhere. So data should be more accurately priced as Bits/second/kilometer.

This is a very difficult thing to answer.
For example as Shalmanese pointed out it’s the equipment that costs money.
So do you sell connections based upon the theoretical number of them you can get for a piece of equipment or on the number of actual connections.

Say a switch costs $1000 to buy and operate with X% profit (for a simple example), and it can provide 1000 64K connections. Do you charge $1 per 64K connection or, since you have only sold 500 connections do you charge $2 per connection so you don’t lose money.

I work for a large telco and when regulatory bodies ask for our actual cost in providing these connections to make sure we are not overcharging the competitors, who we must give wholesare rates to, it can be a grey area.

Now take into account that predicting the useful life of the equipment is difficult as well as predicting demand and you have a pretty confusing issue on your hands.
Sorry if this is no help.

(sorry for the wordy sentences too)