Slashdot Mirror
HD Video Could 'Choke the Internet'?
richdun writes "Yahoo! is carrying an AP story explaining how ISPs are worried large streaming videos could 'choke the Internet.' This is used as a yet another reason for tiered pricing for access to content providers." From the article: "Most home Internet use is in brief bursts -- an e-mail here, a Web page there. If people start watching streaming video like they watch TV -- for hours at a time -- that puts a strain on the Internet that it wasn't designed for, ISPs say, and beefing up the Internet's capacity to prevent that will be expensive. To offset that cost, ISPs want to start charging content providers to ensure delivery of large video files, for example."
I own a dedicated server and I have to pay per gig for bandwidth... So I have to ask how is this any different than what is already happening?
Are they just asking for more per gig? Or are they asking for money to flow up a chain (from hosts to network operators)?
Roads essentially have, or have had, the same issues. These are funded by state/federal taxes and/or toll roads or some other per-use charges. Perhaps a model like this could work for the internet too.
Engineering is the art of compromise.
So, the bandwidth providers have finally found an actual reason for wanting to charge content owners for content delivery to the consumer. They still have not figured out that the people who should be paying for more bandwidth are the consumers.
Either way--and I say this all the time when someone raises the issue of network neutrality--the Internet was designed to route around troubled, undesirable routes; should bandwidth providers choose to raise the cost of their lines, the Internet will simply route around them. It's as simple as that.
Colin Dean Go a year without DRM
Maybe I'm just silly, but I'd think that this would have a sort of self-limiting effect, much like supply and demand in economic markets. My logic is that as HD video slows down the internet, the incentive to use the internet to watch this kind of stuff will diminish, thus alleviating the pressure. This balance between availability of bandwidth and demand for it, expressed as "cost," or rather, speed of downloads, would make the problem disappear by forcing usage to level out at a point acceptable to all.
And anyway, isn't there tons of dark fibre around?
Of course, I may be insane, and no, I did not RTFA.
Wasn't multicast (http://www.cisco.com/en/US/products/ps6552/produc ts_ios_technology_home.html) supposed to take care of this?
I agree HDTV on the wire could be a serious problem. But, what I've seen from Comcast (my only experience so far) it appears they're introducing extra compression, and the HDTV of a friend gives a status showing a transfer rate of 6MBs. But, this article shows HDTV needing aroudn 20MBs for streaming. To move to a world of on-demand HDTV for the masses would seem to (as they're claiming) require not only some prioritization of the network, but I would think it would also require a more capable internet, i.e., bigger pipes almost everywhere.
In addition, at my friend's, we found that HDTV streams could grind the house network to a crawl, I don't know if it's related (since it really isn't part of the network, but it is coming in on the same coax). Considering everything I've seen and experienced (hiccups in the picture, sometimes outright halting) I don't think HDTV over the wire is ready for prime time yet.
However, if I were a provider, I would have to consider that all of a sudden even a small percentage of my customers could consume all of my bandwidth and would have to come up with some approach to keep the pipes working.
It's very telling they want to charge the content providers and not the consumers
who would be utilizing this bandwidth (as they charge consumers and providers for their
bandwidth now). Why charge producers?
Also, it's the notorious three: Verizon's top lobbyist, Bell South and AT&T making a statement on how they aren't prepared for people downloading more content.
Gotta love those companies for turning over all subscriber phone records without a peep of protest.
They mentioned a video of Colbert heckling the president. Would they still feel this way if Colbert
praised the president?
On the other side of things, cost per gigabyte of bandwidth has dropped markedly and will continue to fall.
But in the short to mid-term, perhaps a case can be made that consumer demand for bandwidth will reach levels that current subscription fees can't cover. This is a reasonable argument, but there's nothing to this argument that requires these costs be offset by content providers.
Right now I'm getting about a half a MB a second over my cable modem. Maybe it will turn out that there are HD audio applications I really want, that will require greater bandwidth. Fine. I'm the one consuming this bandwidth. So let me shop around and find the cheapest provider of super-broadband.
But there's nothing in this article, and no argument I've yet seen, that gives any clear reason why content providers ought to be the one ponying up to cover these extra bandwidth costs. This whole argument is being made by large incumbent ISPs who are looking to extort content providers. It has nothing to do with charging people for what they consume. Those costs have traditionally be borne by Internet users, and they should continue to be.
If I find out that my ISP is charging content providers a toll to reach me, I'll immediately do everything possible to change ISPs.
On another matter, it's telling that this article quotes nobody who says that this is a bad faith argument. The reporting in this article is either inept or corrupt.
I'm generally "Interesting," "Insightful," and even "Funny" here. What the hell happens to me at parties?
Let's review: The ILECs have been salivating for decades over the idea of becoming "cable companies," and distributing television content over the telephone infrastructure. (They wanted to be able to force customers to go only to their servers, but Judge Harold Greene said, "No, you don't get to control both content and carriage, because you'll abuse that position.") For the past several decades, it has been no secret just how much bandwidth video broadcasting requires, even with compression. It has also been no secret that the broadcasting industry has been moving in fits and starts toward hi-def.
Now here we are on the eve of large-scale HD rollout, and the ILECs are whining that the network backbone may not be able to handle the load. Well, kee-ryst on toast, what the fsck have you been doing the last twenty years? You knew Internet "television" was coming, you knew hi-def was coming, you knew it was going to be a bandwidth hog, you had at least twenty years warning, and you're telling us with a straight face that you didn't prepare for it??
And by the way, who else here is old enough to remember a few years ago when the same ILECs were complaining that all those modem users phoning ISPs were overloading their switches, and wanted to start charging a premium for data calls? My response then was as it is now: Why the hell aren't you building out your network?
Sympathy factor zero, Captain. You either get to work and build out the network like you were supposed to be doing, or stand aside and let the CableCos eat your lunch.
Schwab
Editor, A1-AAA AmeriCaptions
Here in Japan, the ISPs still oversell. But at least they give you the option of how much oversell you get screwed on.
When you buy FTTH service from NTT, they have a high-speed and low-speed option. The HS option is twice the price. However, if you look at the systems, both give you 100mbps over single-mode fiber.
What's the difference?
Well, the HS option has 16 customers per DSLAM; the LS option has 32 per.
As US customers become better educated about their line capabilities, expect more ISPs to cater to their needs. But, you better be prepared to pay for it.
Electricity is metered. Water is metered. Hell, even my trash is metered. What makes you think bandwidth will be any different? People need to be prepared to pay, per MB or GB, if they want quality service.
I'd rather you do it wrong, than for me to have to do it at all.
First, the fiber network that was laid out during the .com boom globally by companies like global crossing currently contains a lot of dark fiber. So that part is cheap.
The capacity of a fiber is easily 10Gb/s per color times 125 colors or 1Tb/s, and a cable is easily 700 fibers, so a total of 1Eb/s. Order of magnitude less for ocean fibers.
*Very* HD is 20Mb/s, so a cable will handle 50 million channels.
Cisco's high end router handles up to 70Tb/s.
Lets take the olympics as a scenario:
You are broadcasting 500 concurrent HD channels at 20Mb/s each channel. This is 10Gb/s.
This fills less than 1% of one fiber in the cable.
Now, Every family member in the house watch their own event, so this is 100Mb/s
The Router handles 70Tb/s, so one router supports 700,000 households. So you need 1 router for Seattle, 1 for London etc.
The only clamp on this whole thing is all the ISP whining about problems and clamping down on bandwidth to try to maximize their revenue.
Like DeBeers and diamonds, it is actually a bandwidth glut, and the ISP's are creating an artificially high price for it by limiting supply.
"Fix it"
This "choking the internet" complaint seems to be a cop-out for the laziness of the ISPs toward getting off their butts and really competing to bring a smooth connection to its subscribers.
Yes, by all means, conduct an ad hominem attack on the ISPs rather than considering that this could possible be a difficult problem. Do you have any concept of how difficult it is to design and engineer a network that can handle all of that data and provide a high level of service to all of the end nodes? Then we have to include the cost of the equipment and maintenance, and factor in the time it takes to actually build the thing.
Consider this scenario... A pair of high-rise apartment buildings go up right next to each other. Each one has 15 apartments per floor and is 15 floors high. This is 225 units per building, and with 2 buildings brings it to 450 units. Now if each unit actually wants 10 Mbps so they can download HD video in a reasonable amount of time, this means that one area needs 2.25 Gbps of bandwidth.
Sure, this is reasonable for a local area network, but this isn't a LAN. Maybe all of the users are hitting the same server, but requesting different files. Since they are different files we can't cache locally, and we need every link (and router and switch), capable of handling that 2.25 Gbps. This is in addition to any other traffic that might be travelling those links or routers. Multiply this by all of the apartment buildings/condos/homes in a small city and you can see the problem.
High-performance networks that can handle all of these things are an active research area because we don't have any good solutions. You can't just magically add another switch to upgrade your service. This is a local solution and doesn't address the entire network, or even the network core. More bandwidth in fiber doesn't solve the problem, it just moves it to the switching/router space. We have a lot of different techniques to help switching, such as optical burst switching, but they are still difficult. For example, in OBS how long do you wait to aggregate the different packets? Too short a time and it isn't efficient enough, too long and the inter-packet delay is too high for real-time audio or video. If you set different time limits based on application type you add the overhead of examining the packets to determine what it is.
So before you accuse the ISPs of being lazy, why don't you come up with a solution that scales globally and doesn't cost a trillion dollars and take 50 years to deploy.
Reading code is like reading the dictionary - you have to read half of it before you can go back and understand it.
I should amend that banks can still outright FAIL, if they're run poorly... the Fed doesn't protect against that. What the Fed provides is, essentially, unlimited cash to cover withdrawals if a bank experiences a run. This costs the bank money, they don't provide the service for free.. I think it's called the 'overnight funds rate', but I'm not entirely positive.
Bandwidth is the size of the pipe. No one is advertising unlimited bandwidth, because there is no such thing as an infinite speed modem. They don't really advertise unlimited data transfer either, because the maximum data transfer would be the (speed of the modem)*(length of the billing period). What they do advertise is that you have unlimited access to the pipe, in the sense that it is always on at the speed you paid for, and you can send stuff accross it 24/7. However, ISPs are running into trouble since they oversold the lines on the assumptions that most people would not use their high speed connecion to transfer huge amounts of data - so that's why they are trying to back down from the unlimited access claims.
Electricity is metered because the cost of producing the electricity is directly proportional to use.
Water is metered for the same reason: the more water that's produced, the more in the way of intermediate resources are needed.
But the cost of bits isn't proportional to the number of bits transferred, it's proportional to the rate, and only because the price of the routing equipment is roughly proportional to the speed of the equipment.
But even that relationship is tenuous at best.
You see, thanks to Moore's law, the price of the equipment needed to handle a given amount of bandwidth should continuously drop over time, which means that eventually the expense of providing that bandwidth should be dominated by two things:
So any ISP which charges based on the amount of bandwidth being consumed is likely overcharging the high-bandwidth subscribers so they can undercharge the low-bandwidth subscribers, which they all should be charged the same flat fee.
And any really smart ISP will build infrastructure that's by design as fast as it can be (thus, they would be running fiber and not copper) because in the end, the cost of building out the infrastructure is almost certainly dominated by the cost of the labor to put it in place, or perhaps the cost of the right-of-ways (which is proportional to the distance, not the bandwidth), and not the equipment itself.
We've known since the 80s that fiber would be the fastest transport medium available, simply based on the fact that light has more bandwidth than any other conventional signal. Any bandwidth provider that has built out infrastructure using anything else is an idiot for doing so.
There are a lot of idiot providers in the U.S., thanks to the typical company's inability or unwillingness to look ahead more than a couple of months.
Use 'slashdot stuff' in the subject line in any email you send me if you want to get past the spam filter.
>>You see, thanks to Moore's law
Not really applicable here. Moore states that the number of transistors will double every 18 months. He states nothing about processor speed, bandwidth, or utilization.
However, let's both agree that the cost of tech is going down. A T1 today costs a lot less than a T1 10 years ago. I remember paying thousands in install fees and hundreds for monthly fees. Costs are dropping.
But, I think we can also agree that the customer demand is rapidly outstripping capabilities. ISPs are not structured to give every customer 100% utilization 24/7. Yes, they sold "unlimited bandwidth". Yes, they sold "always on". However, a lot of the fine print advised customers agianst 100% utilization. They just can't get upstream bandwidth cheap enough to resell to customers and still make a profit.
>>The amount of labor required to run fiber is roughly the same as the amount required to run twisted pair.
That's complete bullshit. I have installed fiber and copper. I have run "house cable" from comms closets to the customers' desktops. I have also been in manholes running cable between buildings. Fiber takes a lot more time to install. You need a lot of expensive, specialized tools to install it. You have to be a lot more anal about QA after the install.
>>The amount of labor required to add routes is the same no matter how fast or slow the links in question are.
That's BS too. OSPF and EIGRP are nice, but not perfect. You have to have people qualified to analyze the network before you upgrade. They have to examine every possible reason for the lack of performance. And, after install, they have to go back to find and fix the next bottleneck.
It isn't as easy as letting MRTG graphs show overutilized lines. You can't just take a OC-48 at 80% utilization and upgrade it to a OC-192. A lot of times, telcos save money by finding low utilization backdoors into overtaxed areas.
Cisco and Juniper are not cheap. Neither are the certified techs who really know how to herd them cats like a mofo.
>>And material cost doesn't vary much with the speed of the link, either.
Yet another misleading statement. The tools neede to diagnose noise on a voice line (i.e. a lineman's handset) are a lot less expensive than the tools needed to diagnose malformed cells on a OC-192.
Furthermore, the techs qualified to operate these tools get paid a *shitload* of money. It is not uncommon for a tech holding a Acterna TestPad to earn 4x what the lineman earns.
On top of that, the more lines you have, the more techs you need. You also need a lot more sophistication in the NOC to predict, diagnose, and reroute around broken lines. When an OC-192 drops, networks reel trying to automatically reroute. Well-paid NOC staff can identify low-priority customers (read, residential ISPs and cable ISPs) and disconnect them to perserve customers who would actually notice (and, more to the point, demand a chargeback for the outage). Sure, you could trust a computer or routing table to do that, but paid staff can do a much better job.
>>And any really smart ISP will build infrastructure that's by design as fast as it can be
No residential ISP will start off by hiring a team of CCIEs to install and configure enterprise-class routers. They start off by installing a few DSLAMs and some Cisco 2600s. They link the whole thing together with stickytape, rust, and T1s. Then, as the customer base grows, they start an endless cycle of upgrades.
It'd be nice to have a network designed from the ground up to provide 100mbps FTTD/FTTC/FTTH. Look at Japan and NTT for an example. The problem with that is that there is no room for the "little fish" in that equation. While a lot of Mom&Pop ISPs are gone, their equipment still serves the same customers. The bills just go to AT&T vice Vicki and Kenniths' ISP and resturant.
>>We've known since the 80s that fiber would be the fastest tran
I'd rather you do it wrong, than for me to have to do it at all.
Well actually it is not "a load ..." it is fact. Any network has a maximum throughput and in the case of large downloads from an ISP to multiple households you can reach the capacity limit of that network quite quickly.
This is a common problem of any network and no matter what you do to increase network bandwidth the users will always fill it up and this equally applies to disk storage as well, so it is no wonder that providers want more money for network usage.
Although I think it is a cheap shot at making money out of something that cannot possibly be fully controlled, a cost to the receiver's hip pocket is normally a good way of controlling excessive use.
The only way of reducing huge downloads is to make it cheaper to purchase the DVD, HD-DVD or BluRay movie (can't see that happening), however there are people who will gladly pay more to download/stream the latest movies (go figure).
There ain't no such thing as proprietary standards only proprietary formats. Standards are by definition open.
Not true. The people using Bittorrent, Usenet, etc... are usually more technically savvy users (1-2% of high speed users I think). If you take a lot of regular Joe blows who sit in front of their TVs for hours drinking beer and having him utilize his/her entire pipe, you'll have a huge problem.
.com burst. Quest was already testing multiple OC-768 (80Gbps) circuits then, and this was in 2001. They knew about the future of streaming video then as they do now, so it's time to get back to the original expansion and stop finding excuses to regulate traffic flow.
That being said, it's not THAT expensive to expand backbones. From what I remember when I worked at Nortel, plans were already underway to expand backbones dramatically before the
"Electricity is metered. Water is metered. Hell, even my trash is metered. What makes you think bandwidth will be any different? People need to be prepared to pay, per MB or GB, if they want quality service."
Well, that's just the thing. They're not even trying to do that, they're trying to extort money out of Google and MS instead.
See, it's one of those cases where everyone sold what they can't possibly deliver, and now they're tripping on each other's lies. Everyone promised "FREE UNLIMITED DSL!!!" based on the idea that, nah, you're not actually gonna use it. They figured that, yes, you're gonna see a web site or two, send a couple of emails, maybe even download a MB or two of short pixelated movies, but that on the whole you wouldn't actually _use_ 99.999% of that capacity.
Unfortunately it turns out it's as unsustainable as promising "FREE UNLIMITED ELECTRICITY!!!" and thinking people won't use more of it.
And the problem isn't just one of wishful thinking and creative marketting, but it's always been an outright lie. E.g., there was always some clause hidden in the fine print, or not even there, saying they can kick you out if you use "too much" of that "unlimited" thing you've bought. And for a while it worked to villify those who actually use the unlimited bandwidth they bought, and present them as some predators leeching off the rest of the society, because there were few of them, and everyone else didn't give a rat's arse.
But now it's more and more of them, and there's increasing resistance to buying "FREE UNLIMITED DSL" and then being treated like some kind of heinous criminal if you actually use what you've bought. It worked when those "villains" were some lone nerds running a server at home, but it gets people writing to the relevant authorities when their mom gets mis-treated for spending too much time talking to them on VOIP. Or when they themselves get a nasty letter because little Billy played too much World Of Warcraft. (But more likely, they don't even know why. It just says you've used too much bandwidth.)
No matter how you want to look at it, it's a scam. I'm not even opposed to mettered access as such, but I _am_ opposed to selling something and then villifying the people who use just what they've bought. If they sell something as unlimited, then it damn better be just that. It's like selling monthly bus cards on the explicit claim that you can ride the bus as often as you want to with that card, and then tarring and feathering some retired grandma for riding the bus 6 times a day instead of the 2 times a day your marketroids estimated when they priced that card. It's that sick and dishonest.
And the problem is that now getting out of that losing proposition is a bit of a prisoner's dillema, except the losing move is to confess the truth. Anyone trying to sell a service honestly, a la "ok, guys, it costs X dollars per gigabyte" is losing their customers to those promising "FREE UNLIMITED DSL!!!"
So now the plan is basically "I know!!! Google has money, right? Let's extort some protection money out of Google instead." The ISPs would now like to have their cake and eat it. They'd like to continue to scream "FREE UNLIMITED DSL!!!" all over the place, but be allowed to extort someone else to pay the bill. That's all.
It's not even that Google's search even costs the ISPs that much bandwidth. FFS, it's a simple text page, with no graphics other than the "Gooooogle" letters. Even the Google ads are actually using _much_ less bandwidth than the more traditional ads, which in the meantime have inflated to be hideously huge animated popups or overlays. And certainly Google isn't responsible for P2P file swaps and P2P VOIP traffic.
But Google has money, and the ISP would like to be legally allowed to extort some money from Google. And for that matter, from everyone else doing any business on the Internet.
And the stupidity of it all is that all those sites already paid per gigabyte to their uplink. Having to pay extra so the users of some ISP can see your site -- or can see it without it taking 5 minutes to load -- is nothing short of extortion.
A polar bear is a cartesian bear after a coordinate transform.
What the hell are we paying for if not infrastructure improvements? This isn't water, or electricity where you have to generate the resource - unlike resources, marginal cost of an extra bit is trivial. You just need to make sure your pipes are big enough. And lets not forget that the pipes are getting cheaper.
The arguments about the cost of ISP (or university) upstream feeds for videos or other large files moved around by P2P can be taken care of by making sure the P2P users get most of their data from within the same ISP (or university.) Napster was able to do this early on when it started getting complaints from universities that had big fat LAN connections but relatively small outside connections, but it could do it because it had a relatively centralized database it could use to control neighbor connections, and of course preferring short ping times helped.
BitTorrent doesn't really do that, but it does try to use faster connections when possible. This has somewhat the same effect, though it's much more pronounced for universities than for big ISPs, which have big fat fast pipes that are bigger than they want to pay for. Sounds like there's an obvious market for the telcos to pay Bram to tweak his algorithms some more...
The other scalability tool, which can help for broadcast-style TV, is multicast. Most ISPs could just turn it on if they wanted, but they don't have a good business model for the stuff, much less one that supports peering multicast with other ISPs, and most of the obvious uses look like something that people would pay money for, so you mainly see it inside private networks. Think about the scalability of 10,000 households behind one small-medium telco office watching HDTV at primetime. That's about 9 Mbps per user, which is ok on the line side if you've got the right flavors of DSL, but that's almost 100 Gbps of upstream even though most of the people are watching the same thing. If the telco feeds a multicast down to their office, a Gig Ether can handle about 200 channels and then split it out to the individual subscribers. Sure, the telcos would like to control content so they can charge subscribers more money and compete against the cable TV companies, but a lot of the net neutrality nonsense has been because telco officials are doing the regulatory bonehead thing instead of talking about the real technical issues.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
it is what you get. from your computer/router to the nearest company equipment...
I checked what my ISP is offering and (in rough slovak to english translation): "broadband internet access at speeds up to XXXX/YYY".
So at least I'm not buying connectivity to their network (which I do not want) but connectivity to the Internet (which I do want).
On the other hand, while for now I do not know what is the stance of my ISP on that "tiered thing", they at least do not promise exactly the maximum speeds so essentialy I can get even say 1 bps. As long as it's not fault of my ISP I can handle that (e.g. target server is down or its connection is overloaded or does not match mine in terms of speed).
But the monent I start to experience "underperformance" of my connection and I will be able to reasonably attribute that to "bad" ISP's network/interconnection (e.g. I know the parameters of target server to be fully able to saturate my connection but it did not) I'm going to look for another ISP or for another means of getting connected to the Internet (local networks connecting bunch of private citizens with much better bargaining position with ISPs, etc.).
hany
I was discussing this with a colleague, who insisted it could be done by combining streaming protocols with a swarming protocol like BitTorrent. I was skeptical, and pointed to the lack of success of multicast protocols as indicative that the technology to stream to large numbers of consumers already existed, but wasn't supported by the ISPs or by client software.
After thinking about it, I realized he was right. Multicasting will never work due to apathy of the ISPs, so it will have to be built into the application. Take a HD stream, and introduce a fixed delay that would be acceptable to consumers -- such as 10 minutes. Begin a swarming protocol like BitTorrent, but with a statistical weighting so that packets near the beginning of the stream earn a higher priority than those near the end (of the 10 minute window.)
In theory (according to some back of the envelope queueing theory calculations), it should be possible to ensure that 97% of the packets are there within 10 minutes with an average swarm size and typical xDSL bandwidth -- and if you're running a lossy protocol based on UDP, it won't matter too much about the occasional artefact occurring in the stream if the client player interpolates well.
The benefits of this approach for media providers is if they use a signing system with closed source client (both for Windows and Linux), then they could introduce non-skippable adverts and limited DRM, whilst also saving hugely on bandwidth by leveraging from BitTorrent's advantages.
I hereby release the above idea into the Public Domain, but retain the right to be credited as its originator (unless someone can demonstrate prior art.)
Paul Gillingwater
MBA, CISSP, CISM
I heard the same thing as well while working at nortel :-))
:->
I got the impression they were saying this so people would believe it, so people would buy the equipment needed to light up that dark fibre (which Nortel did a very nice line in
when someone is telling you need something (which you will need because all your competitors will buy it, so why not be ahead of the game) then you have to take it with a pinch of salt when they are also selling it
"The weirdest thing about a mind, is that every answer that you find, is the basis of a brand new cliche" -
As long as the high download users are a relativly small portion of the population the grannies, no problem. IPTV (especially high definition) would cause all those emailing grannies (and lots of other folks) to suddenly become high bandwidth users, too.
It sure seems like mass media television type programming is best suited for a broadcast rather than point to point netword.
Degaussing scares the bad magnetism out of the monitor and fills it with good karma.