| ISP Economics: Just what does a gigabyte cost? Alrighty, I hope this will be the definitive thread, and we can try to answer some questions:
1. Just how do you calculate the cost of a gigabyte of transfer for an ISP?
2. If an ISP were asked by the CRTC to calculate it, what should the CRTC be asking for?
2a) For point-to-point GAS cost (like Bell's service that connects my house to 151 Front)
2b)Actual internet backbone provider cost.
4) Feel free to speculate on what kind of ridiculous math the Bells/Cogecos/Rogers of the world would have to use to come up with stratospheric numbers like $1.50/gigabyte.
said by Krispy:I personally have no interest debating economics with you on my limited time during the week. Start a thread about ISP economics and PM me the link and I might participate on the weekend when I'm not so busy. |
|
| Under 3-cents |
|
 CanerisErikCanerisPremium,VIP
join:2007-10-03
Toronto, ON
kudos:2 | reply to Jaser
said by Jaser :
Just how do you calculate the cost of a gigabyte of transfer for an ISP?
You don't. You do 95%ile 
--
Erik - Caneris - Internet solutions and more. |
|
 otty
join:2008-10-24
Revelstoke, BC | reply to its been done
It's a bit more complicated than that (although probably not a lot more than that). The problem is that most of those costs are proprietary confidential information of the company so they won't be made public. That's what they'll argue with the CRTC.
See this thread for details »Re: Competition- Cogeco just went usage billing |
|
| The costs have been stated. Published and made available.
Maybe many haven't saw it.
Under 3-cents/gig per last years analysis.
Posted on this site and on p2pnet.net
Please don't make me search after a few Friday drinks |
|
|
|
otty
join:2008-10-24
Revelstoke, BC | The problem is that the costs within the network are hard to calculate. You could put a per-GB price on it although that is kind of silly. That is what they these companies are doing for the most part. It's clear how much actual internet costs are for ISPs - very cheap. It's not clear how much a network costs to maintian and develop. These are numbers that are confidential and diffucult to put a per-GB price on. THis is what they are arguing they need to charge for.
I think it's clear they are profitable enough as it is. |
|
| reply to been done
That's internet transit, not metro/regional backhaul from a CO to downtown. Those numbers would be different. |
|
 JGROCKYPremium,VIP
join:2005-05-19
Chatham, ON | reply to its been done
said by its been done :
Under 3-cents
You take whatever the cost per Megabit per second average is and divide it by about 320 and it will give you an estimated cost per Gigabyte.
In our case you're talking about $0.03 per Gigabyte as is. This means the cost per Gigabyte against our transit costs. This assume the difference in other services being costed elsewhere.
When you consider the average usage you can ballpark, on the high side $2-3 more per user in collateral COGs (cost of goods), so we're talking about another $0.05-$0.06 as a worst case scenario...
So, ballpark of worst case scenario on things not costed out elsewhere (ie: the initial service monthly/fee) in the rang of $0.10 to $0.15/GB. This considers Internet traffic, not non-internet traffic as this is not a related or equivalent/apples-to-apples topic.
--
TSI Rocky - TekSavvy Solutions Inc.
Authorized TSI employee ( »TekSavvy FAQ »Official support in the forum )
|
|
| reply to otty
That's what I'm arguing for, an unbundling of the base network (ie: what you'd pay if your usage was 0 gigabytes/month, but could in theory use it whenever you wanted), and the variable costs.
This is like electric utilities, natural gas, water, operating a car, almost any good really, where the fixed cost and variable costs are broken down, so users pay for what they actually consume.
So is there any reason at all a smart ISP wouldn't let users opt in to be charged at the 95th percentile of the ISP as a whole? I understand that this would mean the user wouldn't know what they'll get charged next month, but it should be the cheaper way of doing things in the end. I could even envision "bulk data transfer" applications that would dynamically lower their use depending on what the ISP tells them the transfer cost is or when they're near the 95th percentile, or ramp up when they're nowhere near.
Electric utilities are moving toward this system, and indeed, big industrial users already are.
I recall that just prior to the big blackout of 2003, prices spiked so high that a large steel facility in Hamilton actually closed off for the day due to prices.
What was the name of that company that controlled both the supply AND distribution of a good with relatively inflexible demand? Oh right, Enron... |
|
| reply to JGROCKY
said by JGROCKY:said by its been done :
Under 3-cents
You take whatever the cost per Megabit per second average is and divide it by about 320 and it will give you an estimated cost per Gigabyte. In our case you're talking about $0.03 per Gigabyte as is. This means the cost per Gigabyte against our transit costs. This assume the difference in other services being costed elsewhere. When you consider the average usage you can ballpark, on the high side $2-3 more per user in collateral COGs (cost of goods), so we're talking about another $0.05-$0.06 as a worst case scenario... So, ballpark of worst case scenario on things not costed out elsewhere (ie: the initial service monthly/fee) in the rang of $0.10 to $0.15/GB. This considers Internet traffic, not non-internet traffic as this is not a related or equivalent/apples-to-apples topic. Should I contact and ask Charlie Angus to weigh in again?
I'm tempted. |
|
otty
join:2008-10-24
Revelstoke, BC | reply to otty
said by otty:I think it's clear they are profitable enough as it is. Ooops..oh silly me. There is no such thing as profitable enough. It's only enough when the drones have to cut the service because they can't afford it.
That the "free" market...free to abuse as much as you can get away with. |
|
| said by otty:free to abuse as much as you can get away with. U got it |
|
| reply to Jaser
On a long haul, high density transit route, where a backbone carrier is moving multiple 10G wavelengths worth of traffic, the cost is incredibly low.
On a low density metro aggregation system where a single fiber feeder is serving a few hundred customers, the cost is of course higher.
You have to build your network to handle the peaks, not the averages, so when someone is moving traffic at 3am, that bandwidth is essentially "free". Charging on a per Gigabyte basis does not acknowledge that fact. |
|
| said by HeadSpinning:You have to build your network to handle the peaks, not the averages, so when someone is moving traffic at 3am, that bandwidth is essentially "free". Charging on a per Gigabyte basis does not acknowledge that fact. Yes I recall a Bell employee making that comparison. They equated it to electricity.
It costs more during peak hours, so your use is curbed (ie throttled).
After peak hours the cost is reduced or free (ie unthrottled).
I think they also mentioned that transit costs less after 2-am (something like this).
When I first read that it seemed more to me it was a way to reduce their costs w/o adjusting for what they were selling, or over-selling (ie the "peaks" you mention which Bell attributed only to 5% of the users using P2P and use more than 60-gigs/month).
So saturate as much as you can, then kill off some protocols to reduce the peaks, then start saturating as much as you can again till the next protocol (or latency) kill off. |
|
 pstewartPremium,VIP
join:2005-10-12
Peterborough, ON
kudos:1 | reply to HeadSpinning
said by HeadSpinning:On a long haul, high density transit route, where a backbone carrier is moving multiple 10G wavelengths worth of traffic, the cost is incredibly low. On a low density metro aggregation system where a single fiber feeder is serving a few hundred customers, the cost is of course higher. You have to build your network to handle the peaks, not the averages, so when someone is moving traffic at 3am, that bandwidth is essentially "free". Charging on a per Gigabyte basis does not acknowledge that fact. Well put... so there's a number of factors involved in an ISP from a pure cost perspective (as HeadSpinning already knows):
Internet Transit
Colocation/Tenant Costs
Electricity, Taxes, Utilities etc. etc.
Staff (including benefits, taxes, perks etc)
Network Hardware (big $$$)
Service Contracts (support agreements)
Servers (usually LOTS of them)
Sofware Applications
Telephone System / Phone Lines
Dial-Up Lines (if they do dial-up)
Dial-Up Equipment
Wholesale DSL to Bell (if they do wholesale)
Staff training/support
Backup Power (big $$$)
Climate Control (big $$$)
I have about 50 more items to list.. you get the idea....
Then after all those costs, you have profit. Every company needs profit to survive the ups and downs of a traditional business model.
A typical end-user has no idea what it costs to operate an ISP, turn a profit, and have enough margin to reinvest segments and continually keep hardware/software updated.
With all of these factors it makes it very hard to come with an exact per GB charge that is uniform for all providers - it really isn't possible. If you take ISP X that is spending $400k a year on servers and compare that to ISP Z that is spending $50k a year on servers this changes the landscape considerably and this is only one part of the shopping list from above. |
|
 DeadpoolGo Sens GoPremium,VIP
join:2001-03-29
Canada
kudos:17 | pstewart: Well said. |
|
| dont forget that isp have to pay for printer paper if they print stuff
and toner
also they have to pay for staff
so in fact businesses have costs
but perhaps thats getting off topic a bit... |
|
| Excluding internet backbone transit, is there any disagreement with the following statement:
The cost for an ISP is constant regardless of how many gigabytes are transferred over its systems.
The way I'm seeing it, is that once the system is built, it costs the same to operate whether there's 1 gigabyte or 100000 gigabytes transferred over it.
(I'll just leave the data transferred as a fixed quantity rather than per second/day/month to simplify things)
If the system is only capable of 50000 gigabytes of transfer, then there will be just 50000 transferred, even if 100000 gigabytes "wants" to be transferred.
If it is decided to upgrade the system to transfer 100000 gigabytes, then couldn't we easily figure out the costs of that upgrade to figure out how much should be passed on to the consumer?
Are there any variable costs related to transfer for an ISP at all, other than backbone "on-ramp" bandwidth?
Example: If ISP-X users as a whole used twice as much bandwidth this month, and the system were capable of supporting this, would ISP-X Corp incur any increased costs other than backbone bandwidth?
The way I'm seeing it, the photons and electrons that are used in transfer are free and 100% renewable. Gigabytes aren't like burning coal to make electricity where there's something lost forever (relatively) if we need to generate more kilowatt-hours.
All I can think of is that we might want to apportion the costs of those upgrades to the users that necessitated it, but it seems to me like the costs should be easy to determine. Surely ISPs do cost studies and budgets for major upgrades, right?
Any backbone bandwidth cost increases should be even easier to calculate and pass on. |
|
| reply to Jaser
Histogram of Traffic Usage & Costs |
On this graph, we have 4 scenarios.
1. The red graph depicts "normal" internet usage across a network.
2. The blue graph depicts a scam of 95%ile billing, where usage is LOW most of the time, but peaks to infinity for 4.99% of the day, so the bill at the end is actually very low. We can call this Happy Hour ISP Co. Unless the capacity of the link is infinite, I don't think it would work so well, but it is an interesting case.
3. The green graph depicts the effects of a bandwidth throttle, associated cost savings and increased off-peak bandwidth use. If it is necessary, it is effective.
4. The horizontal yellow line depicts two things:
a) The maximum capacity of the link; and
b) The most cost effective use of the link. The most cost effective use is to utilize the link to 100% utilization 100% of the time. In this scenario, we transferred SIGNIFICANTLY more data than in "normal usage" case 1, but the price is only going to be a few percent more. Cost per gigabyte would be lowest here. |
|
| reply to Jaser
said by Jaser :
Excluding internet backbone transit, is there any disagreement with the following statement:
The cost for an ISP is constant regardless of how many gigabytes are transferred over its systems.
It would probably be better stated as follows:
"The cost for an ISP is constant regardless of how many gigabytes are transferred over its systems up to the maximum capacity of the network as built."
The problem is that the network is built with the assumption that over subscription will always occur.
Networks have to be expanded to handle increasing traffic demands from existing users, as well as new users. Fortunately, the cost per megabit/second of transmission and switching gear is coming down at an incredible rate.
Bell's UBB for wholesale ISPs is of course a cash grab and an attempt to keep customers from leaving their service. The money they would collect from UBB purportedly is to offset the costs of adding capacity to a network that they've already stated is not going to be upgraded, while at the same time denying us access to the new higher capacity network (supposedly "NGN") that by its design is far cheaper to build and expand on a cost per megabit/second basis.
Bell wouldn't be even considering pushing IPTV down that network if this wasn't the case.
The clear fact is that all dominant ISPs are now going towards a Usage Based Billing model. They're doing it to try to stop the video content providers (TV Networks, etc.) from making an end run around their video distribution business (Cable TV, Satellite TV, Telco IPTV) by using a direct-to-consumer streaming model over the Internet.
They can't effectively do this without forcing Independent ISPs to do the same.
Its all about video. |
|
