Tech is too cheap to meter, It's time to manage for Abundance(wired.com)
wired.com
Tech is too cheap to meter, It's time to manage for Abundance
http://www.wired.com/techbiz/it/magazine/17-07/mf_freer?currentPage=1
15 comments
Shouldn't SMS messages already be in the "too cheap to meter" category?
It's my understanding that they started out that way in Europe, which is how they got popular. But then somehow the carriers started charging for them. I don't understand how this happened, and I suspect it warrants an antitrust investigation — surely any carrier that offered free SMS with a slightly higher per-minute voice rate would attract a lot of customers who made voice calls once every month or two, and clean up. So I suspect price-fixing.
It would take a lot of voice calls to make up for charging 10p/10c for a text which costs nothing to send!
I'm not saying it would be more profitable; I'm saying it wouldn't lose money, and a lot of people would prefer that service to the service they have. So in a competitive market, someone would be offering it.
Notice how everything which is technically "too cheap to meter" gets metered anyway. Text messages, ring tones, etc. are all examples of products with a cost of production approaching zero, which consumers are charged for simply because it is possible to do so.
Each of those cases relies upon monopolies, laws against reverse-engineering or both. In a competitive market, cell phones that allowed consumers load sounds from their personal computers as ring tones would quickly dominate.
Except in the case where they're much more expensive, which is the case currently. Smartphones can generally do custom ringtones, right?
I don't think they can be more expensive except because of subsidies for dumb-phones. It doesn't require any extra hardware for a phone to download software from the web-site of your choice, and only two pins routed to the outside for it to download from your computer.
Not necessarily. My phone company is MetroPCS. CDMA/flash phones only, weak selection, you pay full price for the phone, and coverage outside major cities is poor or non-existent. They expand features and coverage area slowly: it's not good if you want the latest and greatest. They sublet the Verizon network. That's the bad.
edit: I can get out of area coverage, but it requires prepayment and is expensive. However, I only need it once or twice a year.
However, there's no contract, you pay at the beginning of the month so there's no bill, and it's flat rate. I get unlimited local, national and international calling (to about 100 countries), along with unlimited voicemail and text for $56 a month. Not only is it cheap, administration takes me about 5 minutes per month - the time to go to their website, click 'pay my bill', and save/print the receipt page. I've had one serious customer service problem in 5 years, and that was due to a local store trying to slap on an unauthorized service charge after I replaced a damaged phone. I called the company support line and got the money back.
mPCS is now the 6th biggest phone network in the US. I have saved thousands, and they only send me a text message to remind me of my billing amount or when they add a new feature - one or two a month.
edit: I can get out of area coverage, but it requires prepayment and is expensive. However, I only need it once or twice a year.
However, there's no contract, you pay at the beginning of the month so there's no bill, and it's flat rate. I get unlimited local, national and international calling (to about 100 countries), along with unlimited voicemail and text for $56 a month. Not only is it cheap, administration takes me about 5 minutes per month - the time to go to their website, click 'pay my bill', and save/print the receipt page. I've had one serious customer service problem in 5 years, and that was due to a local store trying to slap on an unauthorized service charge after I replaced a damaged phone. I called the company support line and got the money back.
mPCS is now the 6th biggest phone network in the US. I have saved thousands, and they only send me a text message to remind me of my billing amount or when they add a new feature - one or two a month.
Indeed, we have gotten tremendously more efficient at metering things, even things that aren't being charged for.
The idea that something would be more expensive to meter than to provide is clearly not of the digital age.
The idea that something would be more expensive to meter than to provide is clearly not of the digital age.
I think that off-peak SMS is definitely more expensive to meter than to provide.
The marginal cost is storing 160 bytes for a few seconds, and transmitting those bytes over idle network bandwidth, which has zero marginal cost, and then over the air, which does have a nonzero cost: 1 watt * 160 bytes / 12 kilobytes per second ≈ 20 millijoules, about US$3 × 10⁻¹⁰ at 6¢/kWh — ⅓ nanodollar.
The cost of storage is a little more complicated to quantify. Page 72 of the new Warehouse-Scale Computers book has a couple of case studies: one machine with four disks with a $10 757 three-year TCO and one machine with an unspecified number of disks (let's assume four here also) and a $8 702 three-year TCO. That's maybe two terabytes × 36 months for, generously, US$11k. That's US$1.5 × 10⁻¹⁰ per byte-month, or US$9.3 × 10⁻¹⁵ per byte-second. (The cost might be a little higher due to the need to store some of your bytes on old servers, or equivalently to depreciate your server faster at first.)
The cost of metering, by comparison, is to send several messages over the network to update redundant billing records in stable storage, where they must remain at least for the rest of the month. In many cases, to- and from- information is retained for the rest of the month and then printed on the bill. Origin, destination, and timestamp are, say, 32 bytes, and have to be stored for half a month on average, for US$2.4 × 10⁻⁹ (per replica), which is more than ten times the cost of transmitting the original message over the air. But then if they're printed on the bill, they might take up 1/300 or so of a sheet of 8½×11 paper, which costs on the order of 3¢ to print, so printing them costs on the order of 1/100¢, or US$1 × 10⁻⁴.
But if you're just incrementing a counter in stable storage, rather than recording per-SMS information, you don't pay any extra marginal storage cost per message; you just have to pay a database update. Most database servers are actually bottlenecked on random I/O rather than sequential I/O, but in order to come up with a minimal rather than a maximal cost of a write, let's assume the billing database server is bottlenecked on sequential output at 400 megabytes per second on its four disks, and we need 8 bytes per update. That's half a million SMSes logged per second, presumably on two redundant servers; if we charge the server TCO entirely to the update rate instead of the storage cost, the US$305/month TCO per server comes out to 4.6 × 10⁻¹⁰, which is still more than the cost of sending the original SMS.
However, by far the biggest cost here, other than the cost of printing, is the fixed cost of maintaining the user's account. Just maintaining their name, address, and phone number on a disk in a data center somewhere costs at least 50 bytes, which is about US$7.5 × 10⁻⁹ per month, the cost of 25 SMSs; they probably don't send more than 30 000 SMSs per month, which even in the worst case cost US$9 × 10⁻⁶ of energy to transmit and US$7.2 × 10⁻⁵ to log for the rest of the month. But several times a minute, the phone talks to the tower to let it know it's still in range, doesn't it? And the cost of accepting a credit card payment from the client (once a month, normally) is in the range of 30¢, isn't it?
So, for 30 000 off-peak SMSs in a month:
Keeping a record of the customer's account: ≪1 microdollar
Providing the service: 9 microdollars
Keeping a detailed record of the service provided in order to bill for it: 72 microdollars
Keeping the phone associated with one cell or another: 36 microdollars
Printing out a paper bill: 300 000 microdollars
Mailing the bill (guessing): 1 000 000 microdollars
Accepting credit-card payment: 300 000 microdollars
Handling a five-minute customer-service phone call (guessing): 1 000 000 microdollars
I suspect that on-peak SMS isn't substantially more expensive to provide, but I'm not doing that calculation right now.
The marginal cost is storing 160 bytes for a few seconds, and transmitting those bytes over idle network bandwidth, which has zero marginal cost, and then over the air, which does have a nonzero cost: 1 watt * 160 bytes / 12 kilobytes per second ≈ 20 millijoules, about US$3 × 10⁻¹⁰ at 6¢/kWh — ⅓ nanodollar.
The cost of storage is a little more complicated to quantify. Page 72 of the new Warehouse-Scale Computers book has a couple of case studies: one machine with four disks with a $10 757 three-year TCO and one machine with an unspecified number of disks (let's assume four here also) and a $8 702 three-year TCO. That's maybe two terabytes × 36 months for, generously, US$11k. That's US$1.5 × 10⁻¹⁰ per byte-month, or US$9.3 × 10⁻¹⁵ per byte-second. (The cost might be a little higher due to the need to store some of your bytes on old servers, or equivalently to depreciate your server faster at first.)
The cost of metering, by comparison, is to send several messages over the network to update redundant billing records in stable storage, where they must remain at least for the rest of the month. In many cases, to- and from- information is retained for the rest of the month and then printed on the bill. Origin, destination, and timestamp are, say, 32 bytes, and have to be stored for half a month on average, for US$2.4 × 10⁻⁹ (per replica), which is more than ten times the cost of transmitting the original message over the air. But then if they're printed on the bill, they might take up 1/300 or so of a sheet of 8½×11 paper, which costs on the order of 3¢ to print, so printing them costs on the order of 1/100¢, or US$1 × 10⁻⁴.
But if you're just incrementing a counter in stable storage, rather than recording per-SMS information, you don't pay any extra marginal storage cost per message; you just have to pay a database update. Most database servers are actually bottlenecked on random I/O rather than sequential I/O, but in order to come up with a minimal rather than a maximal cost of a write, let's assume the billing database server is bottlenecked on sequential output at 400 megabytes per second on its four disks, and we need 8 bytes per update. That's half a million SMSes logged per second, presumably on two redundant servers; if we charge the server TCO entirely to the update rate instead of the storage cost, the US$305/month TCO per server comes out to 4.6 × 10⁻¹⁰, which is still more than the cost of sending the original SMS.
However, by far the biggest cost here, other than the cost of printing, is the fixed cost of maintaining the user's account. Just maintaining their name, address, and phone number on a disk in a data center somewhere costs at least 50 bytes, which is about US$7.5 × 10⁻⁹ per month, the cost of 25 SMSs; they probably don't send more than 30 000 SMSs per month, which even in the worst case cost US$9 × 10⁻⁶ of energy to transmit and US$7.2 × 10⁻⁵ to log for the rest of the month. But several times a minute, the phone talks to the tower to let it know it's still in range, doesn't it? And the cost of accepting a credit card payment from the client (once a month, normally) is in the range of 30¢, isn't it?
So, for 30 000 off-peak SMSs in a month:
Keeping a record of the customer's account: ≪1 microdollar
Providing the service: 9 microdollars
Keeping a detailed record of the service provided in order to bill for it: 72 microdollars
Keeping the phone associated with one cell or another: 36 microdollars
Printing out a paper bill: 300 000 microdollars
Mailing the bill (guessing): 1 000 000 microdollars
Accepting credit-card payment: 300 000 microdollars
Handling a five-minute customer-service phone call (guessing): 1 000 000 microdollars
I suspect that on-peak SMS isn't substantially more expensive to provide, but I'm not doing that calculation right now.
The over the air part of SMS can be zero cost. The SMS message is in a spare 128byte segment of a standard phone-base status message (it can send a new packet specificaly to transmit the SMS but it doesn't have to if you don't mind a few mins latency). If the receiver is on the same base station -like kids texting each other across the classroom it doesn't use any of the phone companies network bandwidth.
It takes more bandwidth to send a message back to HQ to bill for the text message than it does to transmit the SMS.
It takes more bandwidth to send a message back to HQ to bill for the text message than it does to transmit the SMS.
Nice! I didn't know about that; I don't know much about cellular protocols.
> This is the power of waste. When scarce resources become abundant, smart people treat them differently, exploiting them rather than conserving them. It feels wrong, but done right it can change the world.
Try calling it experimentation or play. This dramatic word usage reminds me of how programmers are supposedly "lazy", when we are actually just being efficient by amortizing our development costs.
Try calling it experimentation or play. This dramatic word usage reminds me of how programmers are supposedly "lazy", when we are actually just being efficient by amortizing our development costs.
When free evening/weekend calls were first introduced there were stories of people using their cell phones as baby monitors - the phone co had to introduce a one hour maximum call length.