Let's Be Honest About Bandwidth Rationing
from the misleading-doesn't-help dept
Earlier this week, GigaOM released a new paper on bandwidth caps by Muayyad Al-Chalabi. It’s been getting a fair amount of positive attention from around the web. I have to confess that I find this somewhat mystifying — I can only assume that those linking to it approvingly haven’t actually bothered to read it. Mr. Malik does a lot of good work, but this paper is beneath his usually high standards. It is at times confused, at others dishonest, and almost uniformly irrelevant.
The paper even starts off on the wrong foot with the standard, formulaic, credential-establishing pseudo-academic gibberish, in this case about scale-free networks. Al-Chalabi sagely notes that: “A scale-free network is a network whose degree distribution follows a power law, at least asymptotically. That is, the fraction P(k) of nodes in the network having k connections to other nodes goes for large values of k as P(k) ~ k-y where k is a constant whose value is typically in the range 2<k<3.” Fascinating! But it does sound vaguely familiar… hmm… now where could I have read about scale-free networks before? Oh right! Wikipedia!
There’s no shame in using Wikipedia as a reference (although there probably should be some shame in copying from it verbatim without a citation). No, the real problem is that there’s no indication that the author understands what any of this means, aside, perhaps, from it having something to do with graph theory. Still, at least the information on Wikipedia is correct. In the paper, the end of this section is the point at which things begin to go downhill.
Al-Chalabi’s basic thesis is that the power users likely to be affected by Comcast’s 250GB monthly bandwidth cap are vital to the network’s overall health. To establish this, he talks about Skype — it’s questionable whether a VoIP app that uses sub-dialup amounts of bandwidth is germane to a discussion of high bandwidth use, but let’s press on. He mentions that machines on the Skype network can be classified as nodes, login servers or supernodes. This is true. Supernodes “exhibit higher degrees of connectivity than ordinary nodes”. Also true! But here the paper conflates “connectivity” and “bandwidth use” and everything goes badly wrong. Supernodes sound great — they’re super, after all! — but they’re really just nodes that aren’t sitting behind firewalls. That lets them help with NAT traversal and other network tasks, but it adds up to very little extra bandwidth use. Bandwidth limits are irrelevant to supernodes’ existence.
Al-Chalabi goes on to defend the utility of power users elsewhere, namechecking “web-based applications and networks”, “critical social network hubs” (p. 6) and Hulu, Netflix and iTunes (p. 7). But none of these applications rely on power users to relay their content through the network — they’re all standard client/server apps. In truth, Al-Chalabi’s power users are critical for the health of only one major type of network application: P2P. And that’s fine! P2P is important and has plenty of legitimate uses. But despite the dire predictions of bandwidth-capped startups smothered in their metaphorical cribs, the bandwidth hogs currently on the rise are non-P2P ventures backed by huge corporations. Let’s not pretend that we need to subsidize power users in order to maintain the health of our plucky web startups and their apps. It may well be that someday legitimate P2P-based apps like Joost and Steam arise as a meaningful force at odds with bandwidth limitations. At the moment, though, the initial wave of enthusiasm for integrating P2P into mainstream applications seems to be somewhat stalled.
The most startlingly dishonest part of the whitepaper comes at its end, when Al-Chalabi uses an estimate of monthly data use by an average household in 2012: 200GB per month. He then picks Time Warner’s proposed tiered pricing scheme and uses it to arrive at a monthly bill of over $200. This is ridiculous. It’s quite obvious that bandwidth caps and pricing schemes will adapt as data demands grow, provider networks improve and competition forces carriers to adapt to changing customer demands. I might as well assert that a minute of trans-Atlantic telephone communication costs $300 — that’d be the inflation-adjusted price, based on how much it cost when the service first became available. It’s also odd that Al-Chalabi uses Time Warner’s pricing scheme after primarily discussing Comcast in the rest of the paper. But then, his 2012 bandwidth use estimate falls underneath the 250GB cap that Comcast is instating. Even if the cap didn’t go up over time, your data bill would (theoretically) remain constant throughout his example.
I’m sympathetic to where Al-Chalabi is coming from, and others here at Techdirt are even more so — Mike is on record as opposing bandwidth caps. But this whitepaper merely amounts to a complaint that a free lunch is ending. Bandwidth is clearly an increasingly limited resource. And in capitalist societies, money is how we allocate limited resources. The alternate solutions that Al-Chalabi proposes to the carriers on pages 6 and 8 — like P2P mirrors, improved service and “leveraging… existing relationships with content providers” — either assume that network improvements are free, would gut network neutrality, or are simply nonsense.
Yes, Comcast’s bandwidth cap is a drag. Instead of disconnection, there should be reasonable fees imposed for overages. They should come up with a schedule defining how the cap will increase in the future. And the paper’s suggestion of loosened limits during off-peak times is a good one. But the establishment of an actual, known limit does constitute a real improvement over the company’s frankly despicable past treatment of the issue and its customers. Hopefully the data carriers’ pricing schemes will continue to evolve in a more nuanced direction. Reasonable people can disagree about the specifics, but as this whitepaper accidentally proves, it’s hard to make an honest case that people shouldn’t have to pay for what they use.
Filed Under: bandwidth caps, bandwidth rationing
Comments on “Let's Be Honest About Bandwidth Rationing”
You had me up until...
You had me up until “Bandwidth is clearly an increasingly limited resource.”
You make a statement of opinion as if it were a plain as day fact. While its patently obvious that bandwidth is a finite resource, I see no reason to draw the conclusion that it is increasingly limited. The only case where it is truly “increasingly limited” is in an area serviced only by DSL with no plans for a cable/fiber layout. Any HFC or FTTH network has plenty of scalability in it for quite a very long time. It’s just a matter of the ISP willing to install the gear at the nodes.
Re: You had me up until...
The limitation isn’t the last mile – the link to your house, but the middle miles – the link from the “internet” to the box that serves your house.
All the ISPs have this problem is various sizes – even FIOS has a relatively small middle pipe, but they don’t have enough of a take rate in any one neighborhood for it to be noticeable yet.
The problem is that increasing the size of this middle link is expensive – often a forklift upgrade, with replacement of the entire unit.
This is independent of the source of the content – P2P and cached/central server architectures have the same impact. All P2P does is shift the costs from the content source to the ISP.
Re: You had me up until...
It’s just a matter of the ISP willing to install the gear at the nodes.
“Willingness” in this case isn’t just about them getting off their lazy butts — it’s about paying to do it. That’s what makes bandwidth limited.
Is it possible...?
That Al-Chalabi is the one who wrote the Wikipedia article and thus didn’t want to re-write what he’s already written. It would explain the fact that it didn’t cite it since you don’t have to cite yourself.
Re: Is it possible...?
I suppose it’s possible that Al-Chalabi contributed to the article — it’s had many of contributors — but the original version is by David Wheeler. Consequently a citation is called for even if Al-Chalabi had something to do with the article’s revisions and additions.
But I doubt he did. I’d encourage you to read the relevant section of the whitepaper — the concepts discussed by the Wikipedia text are not expanded upon or placed into context. There’s also the fact that various other parts of the Wikipedia article are rewritten into the section (a reference to Albert-László Barabási, and a John Robb paper from the references section that discusses terrorist networks).
Mental Transaction Costs
Btw, to clarify why I think bandwidth caps and metering are bad, is that they increase mental transaction costs, and those are significant.
For evidence, all you need to do is look at other services that have switched between limited and unlimited usage. The internet itself only became mainstream when AOL shifted from its per hour fee to unlimited. Mobile phone usage only caught on when the plans included huge buckets that almost no one would surpass, rather than charging per minute.
The power of unlimited usage to drive usage and drive innovation shouldn’t be underestimated.
Re: Mental Transaction Costs
I certainly agree that this is a reasonable position, but I do disagree about the significance of those transaction costs. There are many services that have succeeded despite relying upon metered sale of quantities that aren’t easy to keep track of — consider electricity or gas use. Or the mobile industry in non-US parts of the world.
I agree that having a basic initial bucket of data has utility for customers. But I disagree that a one-size-fits-all, all-you-can-eat billing model is appropriate. It certainly isn’t very fair.
Re: Mental Transaction Costs
I agree with your overall point, but I’m not sure that analogy’s totally accurate. Here in Britain (and as far as I know the rest of Western Europe) pre-pay SIM cards -referred to as pay-as-you-go here, a term used in adverts run by British pioneers Vodaphone that became generic as other providers got in on the act- were what really drove the mobile phone to its central place in everyday life; it’s more expensive per minute, but it works out cheaper for a less frequent user.
Actually, now I come to think of it, the fact that we have twice as many cellphone providers for a third of the population of the United States might have been what made the real difference.
You say that “Bandwidth is clearly an increasingly limited resource.” But you don’t back that statement up with data.
It may be that the “limitations” placed on this “resource” are completely arbitrary and artificially created.
The cost of increasing the amount of bandwidth available for everyone is nowhere near the amount that telecoms stand to profit from this artificial “bandwidth shortage” that they’re starting to cry about.
It has nothing to do with a physical limit and everything to do with a desire to blow up profits once again.
It’s very similar in fact, to the reason there so little public R&D money going into solar and wind energy. If we could power our homes from a windmill or panels on the roof, then no big company makes money, and we can’t have that now, can we?
I agree totally… the so-called “bandwidth shortage” is created by broadband ISPs like Comcast being too greedy, unwilling to invest in infrastructure improvements to handle its increased customer base.
Also, as far as terminating people who exceed the cap vs. imposing higher fees, which is, administratively, cheaper? Termiantion, of course. Now, rather than having to figure out extra fees for those who exceed the cap, you just unilaterally cut them off, including any VoIP (Vonage, CallVantaqe, etc.) and audio/video streaming services (Netflix, Apple TV), all of which compete with Comcast’s core business. Interestingly, though, if your internet is cut off because of exceeding the cap, Comcast’s own Digital Voice service will NOT be terminated. Interesting, eh?
The regulators continue to let Comcast and the others get away with it.
Yet again; Right-of-Way dependacy not mentioned.
Sorry but any flappings about broadband that neglect to address right-of-way has very little merit. How this major factor is dubiously left out of the discussion is quite remarkable.
Lack of competition is also a frequently missing aspect to rationalizations for gouging.
I agree that limitations are bad for everybody but good god, this article made my brain hurt
I was planning to add another comment to the GigaOM post but was struggling to do this without writing an essay – now I can just link to this post!
The only thing I am uncomfortable with in your post is the Wikipedia thing. As you’ve demonstrated, there are enough holes in the paper as it is. Many of them come down to hasty preparation and I’m willing to accept that this was an unintentional over site.
The bottom line with reverting to some variant of metered usage is that some people will pay more. I don’t have a problem with that. The real concern is how to avoid this morphing into ‘most people pay more’?
I don’t buy the argument that the “Mental Transaction Costs” are prohibitive – this says something negative about the capabilities of your average Netizen.
So I am in favor of some kind of user pays principle being applied to ‘broadband’ but I do buy the argument that in a market with little competition, there has to be some way to keep the bastards honest.
Just to clarify, there are two Lees contributing to Techdirt. This one was written by Tom, not Tim.
Tom or whoever is reading this post, I think you've been mislead by talk of "free lunch" and capitalism and greed
The reason Time Warner and comcast are doing this is great , pure and simple. Most countries in Europe and Japan don’t have caps, even when there was a Japanese cap it was 1 terabyte a month.
Paying $40-50 a month means that there is no free lunch. In fact caps are probably used for cable cos to make more money, or to avoid investing in bandwidth, or because they can’t support speeds without the caps.
A transmission is a transmission, its based on network capacity, this isn’t like using an electric bill, or a gas bill, does the cable co have a pipe that can transmit x numbers of bits yes or no. That’s what’s matters. The goal of caps is to make more money, or to hope that if I only have 100x capacity of bits, lets hope that caps will in such a way free up some capacity with no guarantee (for example let’s say the peak usage is usually at night and people do their normal light surfing in the day for example) , however let’s then suppose that everybody logs in to watch an HD VIDEO at one time. Caps won’t work then because the pipes will be filled.
By the way , use other countries as an example of broadband caps, and the population density argument has been refuted because most hubs are located in the big cities and that’s where most of the broadband penetration in america is already.
Al chalabni may be a bit wrong on his point as social networking doesn’t take up a whole lot of bandwidth, but I respectfully disagree with your opinion TOm and I hope you will comment and try to justify these particular caps again.
Re: Tom or whoever is reading this post, I think you've been mislead by talk of
I feel as though a lot of the comments here amount to people simply asserting that there is no shortage of bandwidth facing ISPs, or that the costs of a full pipe versus an empty one are the same. In a sense this is true — networks can always be upgraded, and peering charges are relatively small. But eventually demand for bandwidth will outstrip a segment’s capacity, and it will have to be replaced or otherwise improved. This costs money, and happens more frequently as your per-customer bandwidth use increases.
The question of whether our ISPs are pricing their services fairly is, to some extent, a separate question. Could Comcast accept a smaller profit margin per customer? Is the lack of broadband competition in most markets leading to higher prices than we’d otherwise have? The answer to both of these questions is almost certainly yes.
That doesn’t mean that they’re acting avariciously or in bad faith when it comes to the caps, though. Consider Comcast’s approach to those who violate their cap (either when it was unstated or now that it’s explicit). They’re not implementing overage charges — they’re simply disconnecting customers who go over the cap. How is this consistent with the idea of bandwidth caps representing a cynical way to extract more money from each customer?
The only way this policy makes sense is if heavy bandwidth users represent a cost to Comcast — whether through peering charges or, more plausibly, through their impact on other customers’ service: the more heavy users, the sooner upgrades have to be paid for in order to maintain an acceptable level of service.
Now, look: Comcast is a lousy company, and so are most of their fellow ISPs. Their service is expensive and customer support is often terrible. They’re in the business of making money, not of making friends. But it should be obvious that 1) their cost per customer is not non-zero and 2) heavy bandwidth users represent a greater cost to them than do light users, primarily due to the network upgrades they make necessary.
In other words, heavy users consume more resources — and if someone consumes more resources, it’s reasonable to ask that they pay more, too. To claim that everyone should pay the same regardless of what they use is unfair; to claim that the heaviest users are somehow doing everyone else a favor, as Al-Chalabi does, is ludicrous.
Re: Re: Tom or whoever is reading this post, I think you've been mislead by talk of
Sorry — “their cost per customer is not non-zero” s/b “their cost per customer is not zero”.
Re: Re: Tom or whoever is reading this post, I think you've been mislead by talk of
In other words, heavy users consume more resources — and if someone consumes more resources, it’s reasonable to ask that they pay more, too.
Sure, but as you say, “They’re not implementing overage charges — they’re simply disconnecting customers who go over the cap.” You also say “the more heavy users, the sooner upgrades have to be paid for in order to maintain an acceptable level of service.” I think that’s the crux of it. They’re only getting $X a month whether you check your email every other day or download HD movies constantly, so if they get rid of the heavy users, 1) their profit margins go up, 2) they don’t have to upgrade their network, keeping costs down, 3) they probably incur fewer administrative costs than would be the case if they adjusted the heavy users’ bills. I think it’s a slimy way to avoid giving customers what they really want. If there were lots of competition, I would say it’s great. Either these caps are plenty big and they would do fine, or they would actually affect customers who would go to the unlimited competition. And they would be willing to pay more for it, too. However, with extremely limited competition in most markets, personally I have a problem with it because there’s no effective way for customers to complain or fight back.
Re: Re: Re: Tom or whoever is reading this post, I think you've been mislead by talk of
I think this is a fair criticism — disconnecting users is not really a sufficient response to the need to discourage high bandwidth use. Overage charges are more appropriate, and something that Comcast considered but rejected, apparently fearing that they would confuse or scare off customers. It makes sense — right now there aren’t great ways for the average user to track their monthly bandwidth use, and getting an unexpectedly large bill is a great way to anger customers.
But other ISPs are implementing overage charges, and I think Comcast probably will, too, as time goes on. Metered bandwidth is where we need to end up (with, perhaps, an initial pool of “free” bandwidth associated with the base monthly fee). Right now the ISPs are afraid that customers will be turned off by that; I think that fear is misplaced. So long as they can resist viewing a switch in pricing schemes as an opportunity to sneakily jack up prices, I think customers would be willing to go along.
Re: Re: Re:2 Tom or whoever is reading this post, I think you've been mislead by talk of
I agree. I wouldn’t have a problem with a bandwidth cap if it’s done well. If it’s a small cap with unlimited or higher cap plans being significantly more expensive, as you say it would be just a way to start charging more. If a big cap is available for a reasonable price, with reasonable overage charges, and I can see how much I’m using, I would be fine with it.
Re: Tom or whoever is reading this post, I think you've been mislead by talk of "free lunch" and capitalism and greed
“a transmission is a transimission, it’s based on network capacity”
first off: electricity is also transmitted, over wires no less. so logically it can be very much like an electric bill.
2nd off, your missing 2 important realizations:
1. most requests placed over a network consist of MULTIPLE transmissions.
2. as 13 pointed out, at some point your sharing a wire w/ one or more other users.
a single transmission equates roughly to one frame, wh/ for ethernet (based on a quick googling) is around 64 bytes (that’s a little over 1/20th of a kibibyte folks, to give you a further idea most simple web pages rank up @ several kibibytes ). Just about any request sent over the network is going to consist of several transmissions/frames.
activity between 2 nodes on the network is a series of frames. if more than one request is pending, then the 2 are interlaced; i.e. for a request A and a request B, the stream of packets could look like this:
A B B A A B …
as #13 pointed out, the problem isn’t on the last mile to your house. you can use the full throughput of that connection as much as you want and never hurt another soul (unless your on a broadband cable connection, wh/ is shared). of course, to contact anything outside of your ISP’s facilities, you have to use their SHARED pipes.
more active user’s essentially are allocated more of the available frame ‘slots’ . so for requests from active users A versus inactive users B, the stream on the shared pipe will tend to looking like this:
A A A A … B A B… A A B …
in short, the active users are getting a bigger slice of the available transmission TIME at the shared resources.
add in the fact that the average network use has increased MUCH faster than network transmision speeds, and you have an environment where users are competing.
hence the idea for things like bandwidth caps, extra charges, etc. to keep things fair; either by preventing power users from hogging all the bandwidth (by wh/ we mean taking all the transmission time on the shared pipes) or making them pay for their above average usage.
Has anyone noted...
The fact that as more and more people turn to sites like Hulu and YouTube for entertainment (especially in college markets), Comcast gets more and more on top of setting bandwidth caps?
Myself and thousands of others no longer even purchase cable television because I can replace it (at some loss of convenience, and some gain) with sites like Hulu. I think Comcast is just jumping early on the potential market loss here, and is nipping it in the butt before they see it as a real problem.
Also, and I’ve said this time and time again, 250gb is NOTHING in bandwidth for a moderate to high user. If they jump me up to over $100 a month for mediocre service I’ll just skip straight up to a T1 and share it with my neighbors. This is outrageous, its not even half my storage drive in bandwidth.
Between class work (large video files, art files, etc.), internet TV, and some web design on the side I’ll have filled my monthly quota, and that’s without any gaming. Basically if these caps and tier systems become standard, the cable company can go stick it you know where and I wont live anywhere without a direct connection.
Instead of disconnection...
Yes, America will have to get used to bandwidth caps like the rest of us, but instead of disconnection or extra charges, ISPs should throttle the account to a lower speed for the rest of the month, like they do here in Australia. We used to just get charged extra when we went over the limit, but that upset a lot of people, so now we don’t get cut off, or charged, we just have to live with a slower connection until the end of the month.
In this article, it says:”the fraction P(k) of nodes in the network having k connections to other nodes goes for large values of k as P(k) ~ k^y where k is a constant whose value is typically in the range 2
Get what you pay for
Here in the UK our ISP’s have been selling access too cheaply for years and are having to introduce caps & throttling as they are not making the money to pay for the number of users they have attracted.
Even those with unlimited download accounts have found that unlimited does have limit and they get kicked off.
Its a matter of greed , not fairness
In Europe, Japan, and South korea they don’t have this problem and the argument of population density doesn’t really apply since most bandwidth hubs are already in major cities.
They don’t have caps, and even if they did it would be far greater than comcast’s caps.
COMCAST IS GREEDY, $40-50 A MONTH IS MORE THAN ENOUGH TO FULFULL DEMAND.
Tom I understand your idea , but I think this is more as a profit gimmick , just look at their cable charges, and the fact that they will do anything to increase money.
Look, how much does access in europe and asia cost for the same speed, or there caps and how much are they?
Just look at text messaging, it doesn’t cost much to send a text message, yet the carries want $20/month or so for unlimited texting.
In many markets, maybe not yours, but mine for sure. Comcast has a monopoly. I can’t get any other form of broadband, save for satelite, and we all know that isn’t much of a choice. Certainly not if one enjoys a latency sensitive gaming experience, whether it be a PC game or console.
Having said that, I pay what I feel is a way too much money for the service I get not because the internet connection is bad so much as because their customer service is horrendous. I can’t count on Comcast at all. It’s a joke to even think for a skinny second that they care about any individual customer at all.
Is it really true that $60/month can’t cover the cost of more than 250 Gig for that time period ?
That’s roughly 8 gig a day right ? That’s around $2 for 8 gig ?
Tom, have you been a Time Warner customer before? “so long as they can resist viewing a switch in pricing schemes as an opportunity to sneakily jack up prices”
I think that is Time Warners mission statement! That is their M.O. in North Texas – increase the price every month until you call to complain about it.
I don’t know the technology, but take AT&T’s U-Verse for example. It claims to offer you, over internet lines, ability to record FOUR HD SHOWS simultaneously. I know a standard def movie takes about 7 gigs of data. If they can move that much data just for TV, how can competitors claim 5GB bandwidth caps (like Time Warner wants) are necessary?
Common sense tells me it’s not.
This week I downloaded Crysis: Warhead/Wars, a couple Netflix streams, some legally downloaded music, plus regular traffic. Crysis was 15 gigs download by itself. I would spit in the face of a Time Warner cap — I’d rather not have internet access and I would cancel the cable as an added “F You” to the company.
Re: Tom Lee
Multicast is more efficient than unicast. Also, all that U-Verse HD video never crosses the Internet.
One more thing Tom
I think you’re missing a main point here as well. All these broadband companies were seeking customers in droves when they realized they could triumph cheaply over dial up. They asked users to come by the millions, all their commercials ragging on dial up and, in the case of cable companies, DSL.
Now they’ve overreached into the market and bandwidth is, apparently, “an issue” so they’re putting it back on the customers that they went after aggressively.
Sounds like bad business to me. If they’re worried about infrastructure I really shouldn’t see a single Comcast Broadband commercial until its fixed. Just MHO
Re: One more thing Tom
Well… yes. They’re going to try to maximize profits and squeeze as many users onto a given network segment as possible. Is this a surprise? I hope not.
None of it changes the fact that resources are constrained and consequently the ISPs must find ways to allocate them efficiently. If competition exists — admittedly a big if — prices will be forced down, providing an optimal allocation of bandwidth for minimal price.
I still haven’t heard a coherent argument for how users can be able to use as much bandwidth as they care to on an indefinite basis. In fact, aside from Mike’s discussion of transaction costs, I haven’t heard an explanation of why they ought to be able to, either.
Constructively critcism with Tom
For off people Pay for Internet access so there was never any free lunch. In some parts of the country , Comcast is the only ISP , or the only alternative is dsl by verizon or some other telco which may not exist at all due to distance issues, certain small telco companies , or very expensive to deploy.
Let’s forget the other countries that price broadband much less that the United states and have no caps at all and even if they do very generous caps.
Let’s focus on ISPS, ISP THEMSELVES PAY FOR BROADBAND BY SPEED NOT USAGE AND WHOLESALE BROADBAND PRICES HAVE DROPPED BY 40% OVER THE LAST YEAR!
Comcast, Tw is just greedy because the are oftentimes the only solution and its as if people have much of a choice with ISPS anymore.
It may happen again.
Besides the point, there was never any free lunch its greed and a way to come up with more money. Consider text messaging , companies want the $20 or so for unlimited text messaging even though it hardly costs them a penny, and they charge 10 cents or so for each message.