|
Bingo Garth -- you hit the nail on the head far better than I ever could have.
Completely incidentally, but for unrelated reasons, we see similar behaviors in the Internet and telco markets, which after some decades is now giving us the problem of net neutrality -- the flat-rate pricing everyone prefers creates a commons which inexorably gets trashed as people try to exploit it en masse. Why? For the same reasons a run on the bank in a fractional-reserve system is disastrous: overselling your reserves.
As a former operator of two ISPs, here's a relevant example, drawn from actual, real-world experience.
Suppose a T1 costs $500/month to maintain. A T1 carries 24 64Kbps channels (of which 56Kbps is usable for customer traffic), so if you want to offer 56Kbps dial-up service, you can only support 24 concurrent callers. So, as a business owner, do you charge people who dial in a value in proportion to their monthly usage? Or, do you offer a flat rate of $21/month to cover T1 expenses used?
It turns out that not everyone will be online all the time, and so you get some people hoarding their connections (e.g., gamers or Youtube surfers) while others use it sporadically (e.g., weather updates, what's on at the movies, etc.). If both are charged the same rate, those which utilize the service the least effectively subsidize those who use the service the most.
Problem: your provider is charging you $500 whether the T1 is actually used or not, so to ensure full utilization of your resources, you sell your product to more customers (who, conveniently, also pay $21/month). Your sporadic users now must compete for bandwidth against more users, leading (eventually) to busy-signals or other related phenomena. That way, you, as a service provider, gets your money's worth; in fact, you're earning a profit hand over fist, because you're actually making more than your expenses paid out every month, by quite a bit I might add. This is apparently a sound business plan.
Except it's not. It is here that the problem arises: natural chaotic laws state that, at some point, large numbers of your users will try to access the network at once. And when that happens, a fraction (all but 24 of them) will get busy signals. Question: can your business absorb the technical support overhead (and, perhaps, the refunds due the customers) and still remain profitable using the flat-rate system? Answer: Yes, as long as you keep your over-sold ratio to about 10:1 (e.g., a 24-channel T1 can reliably support about 240 sporadic users before you feel the crunch; at Armored Internet, we had about 350 customers on a single T1, and tech support calls and refunds had become quite noticable).
You get the same effect with always-on technologies like wireless providers. You pay a flat rate to your cell provider. Yet, the cell carriers have only so much bandwidth to offer in the RF spectrum. They pay a fixed rate to the FCC for license fees, so they over-sell their spectrum because they know that people won't all be online at the same time. Except, it turns out, that people are online with ever-increasing frequency, particularly as more and more services go the way of the Internet. For reasons I'm not aware, the carriers aren't able to acquire more bandwidth (perhaps to preserve their existing infrastructure investments), and so they start charging extra for those who prefer faster service. More precisely, they charge extra for those who desire priority packet routing; those who don't pay the premium will have statistically higher packet loss rates.
But this only affects wireless technologies, right? Nope. This has been a burgeoning issue for wired systems as well. I remember when SBC throttled customer traffic to NNTP behind customers' backs. I remember (and have experienced myself) when Comcast issued TCP RST packets to prevent BitTorrent sessions from consuming larger fractions of a customer's contracted bandwidth. See, the more bandwidth you actually use, the lower the company's profits (since it's predicated on overselling bandwidth used on a sporadic basis, remember?), and so it's in the company's best interests to punish its users, not support them.
You would think this all makes perfect sense, since we're so accustomed to this model of business. We have the benefits of both price discrimination and flat rate pricing! Except, it's not reflective of reality. People should pay for what they actually use/own/etc fairly, and not subsidize others in the process. That's the burgeoning red flag here: you need to ask yourself, "Can the system be used to subsidize others?" If it is, it's a business model that will, at some time in the future, lead to an over-extended commons. The laws of supply, demand, and thermodynamics demand it.
None of this would have happened if people rejected the flat rate payment model, and instead paid a rate concomitant with actual bandwidth use. A sane, sound billing policy would consist of:
* a flat rate the customer agrees to pay to cover business operations. This fee goes towards paying people's wages. As people get hired, this rate goes up. As more customers sign up, this rate goes down. The basic formula here is R=(nP+O)/C, where n is the number of employees, P is their average pay grade, C is the number of customers, and O are the fixed expenses of the business (e.g., rent). It's OK to also add a reasonable fixed profit here too, of course.
* a variable rate based on the length of time a customer utilizes the company's switching fabric. For example, sending a 1500 byte packet on a 56kbps link on that T1 will consume somewhere around 0.2 seconds of time. Out of a 30 day period, this amounts to only $0.0000016. Yeah, it doesn't seem like a lot, but when you consider how long the user actually uses the network, it really does add up to cover that user's share of your expenses. If the user stays connected for the full 30 day period (e.g., as a business might), or uses a solid 30 days worth of traffic (always-on technology like cable or DSL), you've accrued a bill for 1/24th the T1 you're using. Note that as a customer who sends that 1500 byte packet, I will probably end up paying $0.0000032, because of profit. This lets the ISP earn profit with far fewer customers than the flat rate model.
What does this give us? Financial projections will be in line with bandwidth consumption projections, for starters. The heaviest users of the service subsidize infrastructure upgrades. It also lets businesses adequately fund research into providing faster infrastructures, or more advanced core technologies. Observe that spam consists of more than half of the core infrastructure's bandwidth consumption, and yet email still is passed uncompressed!! And speaking of spam, why is it such a problem? Because it's cheap to send massive amounts of data, and relatively expensive to receive it (indeed, the receivers subsidize the senders to a large extent!). If you make the cost of sending e-mail reflect actual resource consumption rates, you'll suddenly find spam rate will drop like a rock, perhaps into insignificance.
Anyway, sorry for the soap-box presentation here. I just find the parallel's between Intel's democratization of warranty pricing (if John's analysis of Intel's behavior proves accurate), Garth's observations, and my own experience from the ISP boom-days to be uncanny. I predict progress in the PC industry will stumble or even falter all-together if this model becomes widespread.
|
Login
Register
FAQ
Search
