Slashdot Mirror
Will Internet TV Crash the Internet?
Stony Stevenson writes "Analyst groups and Cisco have come out saying that the internet is heading for a crash unless it increases its bandwidth capabilities which are being strangled by the increased use of Web TV.
Stan Schatt, research director at ABI said: "Uploading bandwidth is going to have to increase, and the cable providers are going to get killed on bandwidth as HD programming becomes more commonplace." He added that the solution to the problem is to change to digital switching and move to IPTV. "They will be brought kicking and screaming into the 21st century," he said.
Cisco weighed into the argument, adding that it had found American video websites currently transmit more data per month than the entire amount of traffic sent over the internet in 2000."
whether they are going to give us what want, and find a way to stay profitable ... or not. In other words, they're going to have to start acting like real businesses.
The higher the technology, the sharper that two-edged sword.
They certainly have had time to deploy it.
How many times have "experts" predicted the imminent death of the Internet?
Hail Eris, full of mischief...
E pluribus sanguinem
then let's have multicasting. There you are, another good reason for IPv6. Get to it.
Internet providers like Comcast will simply do what they've been doing. They've been throttling bittorrent because of the bandwidth it can take up. They'll simply throttle or block any internet TV that they don't specifically provide since it would be considered competitive to their cable TV offerings.
If you ask me, the whole "problem" is a bunch of balony. ISPs oversubscribe their services, because most people just browse websites, and that's low-bandwidth. Now, they're realising they can't do that, because people are using youtube and bittorrent, and that's about to reach critical mass when people like the BBC legitimize it in a consumer-oriented shrink-wrap. Suddenly, ISPs can't claim that people who actually USE their services are doing something immoral or illegal.
So, what's the problem again? You sold a service extra-cheap, because you didn't think you'd have to provide the full service? Tough. Get real, and sell what we're buying. The prices might go up, sure, but either we'll pay, or we won't care about the new service. Your upstream providers might charge too much for bandwidth, but that'll soon change as ISPs start demanding more.
Choice quotes from this article written at the close of 1995:
http://www.infoworld.com/cgi-bin/displayNew.pl?/mI've spoken with a few engineers in the IPTV business... they're al about multicasting and QOS delivery. I'm going to go out on the limb and say... uhhh.... no. Why?
Because that's NOT what internet TV is all about. Sure, for some content think it's great. Like ABC, Fox, whatever - they can do the multicast. But for the rest of the content providers, it's going to be on-demand. And that solution is really quite simple. And it makes money.
Basically you take an Akamai like model and extend it. Deploy caching servers right to the ISP's - on the customer doorstep. Offer subscriptions to the customers and the ISP gets a chunk of the monthly. Customers get instant access to the content from the caching server. Content people get a chunk from the number of views statistically. ISP's only have to move content over their uplink once for all their customers nearby.
Best part is you could do it securely for the media providers, and give people a reason to use the service (more shows, better quality, faster delivery). Eventually you offer sell-up items like movies, sporting events, etc. In other words it would be better than cable, cheaper than cable, and far cheaper to operate.
There's all kinds of great stuff you could do here - and you could do it on the cheap and make beaucoup bucks. So, ya know... send me a bag of gold hehehe.
I said no... but I missed and it came out yes.
Microsoft OLE DB Provider for ODBC Drivers error '80004005'
/efytimes/fullnews.asp, line 76
[Microsoft][ODBC SQL Server Driver][SQL Server]Transaction (Process ID 238) was deadlocked on lock resources with another process and has been chosen as the deadlock victim. Rerun the transaction.
What's the deal with everybody forgetting the ???? step? You know, if you were working as a CEO, your company would be bankrupt by now.
The internet will never be heading for a "crash", all that will happen is broadband customers will have their packets throttled to whatever limit the upstream provider wants. This has already been happening for almost the last ten years. It's convenient for people who want the broadband providers to upgrade their bandwidth to reference this "crash" idea but it is impossible to ever actually happen due to the traffic shaping already in effect at most (if not all large) ISP's today.
Here in france, you can get 30meg for 30E.
And it's no crappy bandwith.
Because here there is a real competion between internet providers.
The internet is pretty stable even with people uploading and downlonding (up cap is 1meg).
The probleme is that internet service providers in the US and UK don't want spend money to put in fiber optics...
In Japan, most of the people get a fiber to there home... And they get 100meg both ways (not 100% sure..) and they don't have problemes...
The hole internet is going to collapse is FUD. It's only because service providers don't want to evolve.
Available bandwidth is currently deliberately limited by the major incumbents. This manufactured scarcity drives the price up. There is more than enough dark fiber to meet our needs for decades to come.
The incumbents are about to discover that people will only put up with this for so long before they mandate municipal information infrastructure. Fiber is the bridge to the global economy and building bridges is one of the justifications for government exist. If your state and local governments won't do it, mine will - and your kids will find it that much harder to compete with mine.
Fiber is not made of some rare mineral. It is processed sand.
Help stamp out iliturcy.
BitTorrent and eMule could prioritize downloading from people on their ISP's subnet or from people with a low ping/traceroute or the same city.
Live TV could solve its problem with multicasting.
Google/YouTube, I don't know how they can solve problems their model creates.
In a well-known fairy tale a boy enjoyed the attention he got when he cried out there's a wolf in the village! - but after a while, people stopped listening to him, and when there was really a wolf, no-one believed him, and the wolf stole his shoes and socks and his ipod and ran off with them into the forest.
:-), how about it?
The problem with people saying such-and-such will mean the certain end of so-and-so is that, like the boy in the story, they weaken our credulity. What is really meant here is that, if the growth of video downloading continues at the same rate, and no other changes happen, the current system will bog down. And maybe that's true.
I remember a huge thread on Usenet lasting months and months, or so it seemed, Imminent death of the network predicted, and that was in the early 1980s.
Yes, video delivery is something to take seriously. The distinction between downloading a movie for later viewing (I would probably want it to be error-free) and watching streaming video (compression is OK, and I'd want the network to drop packets if I got behind, which is part of what IPv6 quality of service is about) might be part of the solution here. Of course, as people get larger desktop screens with higher resolution, the demand even for static images is increasing. 640x480 doesn't cut it for most people today. And most computer users have stereo sound. Or play games in which network latency is significant. Violent games in which you pretend to be a wolf! And videoconferencing, TV-on-demand (as per original article, e.g. joost), and maybe soon 3d holographic pornography is coming.
The music and video industry would do well to spend a fraction of their current legal bills on researching more efficient delivery. Maybe encouraging deployment of IPv6 multicast, for example, so a single stream can go to thousands of users. Or paid subscription p2p networks. Or cascading servers. For that matter, probably we-who-write-the-standards could help by defining cache protocols that can interoperate with advertising, and can reliably send back access logs, maybe anonymized. Video CEOs, I know you read slashdot
But, shouting "wolves stole my socks" or "the sky is falling" won't help. Although if either of those things does happen, make sure to put the video up on youtube, OK?
Live barefoot!
free engravings/woodcuts
Scaling up bandwidth isn't the final solution, though. As backbones scale up, so do end user connections. You're always going to have bottlenecks. If I'm downloading a DVD over the Internet, that bottleneck is usually my DSL connection. If a thousand people spontaneously decide to download the same DVD (assuming simple unicast), the bottleneck might be at the server's end.
Unless you know of a way to transfer data infinitely fast, these bottlenecks will always exist and will always constrain bandwidth. The trick isn't an arm's race, it's deploying technologies like QoS to allow services to work as anticipated *despite* a congested link. When your 10Gbit SuperDSL line becomes fully utilized (aka congested) with a BitTorrent transfer, you want your 10Mbit IPTV stream to proceed without any packets getting delayed too severely, right? You have to prioritize. That's what QoS is for.
Except the extreme net neutrality crowd doesn't want to allow that.
This problem only exists because some people think the idea of congestion control is evil. (Or people think the idea of providers arbitrarily *degrading* service is evil, but the proposed solution also outlaws legitimate congestion control.) Scaling up bandwidth doesn't solve the problem because certain types of bandwidth-unconstrained services (such as BitTorrent or even a simple HTTP or FTP transfer) will attempt to use as much bandwidth as it can between the hosts. You're going to have a bottleneck (congestion) somewhere. "Infinite bandwidth" is silly. "Lots of idle bandwidth" is stupid and expensive.
Despite this, the FCC says America has the highest broadband deployment rate in the world and President Bush has set a goal of having broadband available to every U.S. home by the end of this year. What have these guys been smoking? Nothing, actually, they simply redefined "broadband" as any Internet service with a download speed of 200 kilobits per second or better. That's less than one percent the target speed set in 1994 that we were supposed to have achieved by 2000 under regulations that still remain in place. This sounds like the telcos' modus operandi to a T. Recall a few years ago, when the FCC eliminated some surcharges, and they continued charging them to customers (as "cost recovery fees") until so many people got angry that the Federal government slapped them down? This is just the same thing, writ even larger.
Although I'm sure there are most corrupt agencies somewhere in the government, I can't think of one that's more bald-facedly corrupt than the FCC. Until we can the whole business and replace it with an organization -- and people -- who have as their mandate the best interests of the citizens of the United States, rather than the telecommunications companies, we're never going to have a first-class communications infrastructure. And the longer we keep the current bunch of bent industry shills and political operatives in place, the worse of a backwater the U.S. will become.
"Ladies and gentlemen, my killbot features Lotus Notes and a machine gun. It is the finest available."
Dark Fiber FAQ
I never said it's a final solution, but at this point with bandwidth being overcommitted 100:1, no amount of QoS will even come close to solving the problem. 6-8 Mbps down is adequate for most uses today but 80Kbps (your approximate share of the ISP's upstream bandwidth) is nowhere close. Today's 1Mbps up is not quite adequate (we might be better off with a 4/5 split). There is no Qos or congestion control that can make 80Kbps adequate even with multi-casting once phone service moves to VoIP (consider a 2 teen family) especially when people will inevitably want a webcam feed to go with it soon enough.
Large scale content-neutral caching at the ISP level would help a lot for other sorts of content (and will STILL be useful even with $9/Mb upstream) but of course, you can't cache a real time videophone conversation.
The one and only reason net-neutrality advocates object to QoS and congestion control is that they are WELL aware that ISPs will use it primarily to double dip and avoid getting more upstream even if the price drops. Further, they will blame poor application performance on any/everything BUT their own craptastic policies and will deny that they even know what QoS is.
Further, since it's much easier to plausibly deny poor bandwidth than an outright null routing, you'll see content the isp doesn't like not quite disappearing (since that would be telling) but becoming so difficult to successfully access that most give up on it.
It's also worth noting that QoS isn't a free lunch either. There isn't a well utilized router anywhere on the net that won't have to be upgraded significantly before it will have the raw power and buffer space needed to actually do QoS. The deeper a packet must be examined, the more silicon you must throw at it. Something has to count those egress tokens and the queued packets have to be put somewhere. There's a reason Cisco is such a big fan of QoS!
A GOOD way to deploy QoS is on the client side. If MY box that *I* have root access on re-prioritizes traffic then I win. Even incoming bandwidth usage can be sorta managed by playing with window size in TCP.
It might be that the problem of ISPs playing dirty with congestion control and QoS can be controlled by forcing them to advertise customer share of the upstream rather than the raw line speed. Where that varies by destination or protocol, they must advertise the smallest allocation. That way if they try to sell fast ethernet to the home but only allocate 2.5Kbps per customer, they are forced to advertise 2.5Kbps service and endure the competition (fairly) comparing them to a mid-'80s dialup service.
Why is it that I hear about ISPs in Japan, Korea, and Europe offering bandwidths up to 100Mb/s for prices under $30/month and in the USA we're still stuck in 1999 pricing and speed-wise? Could it possibly be that the PHBs at the various ISP corporations are deliberately screwing us to avoid actually building out their backbone networks properly?? Just asking...
A team at Cisco decided to build a big router. This was all the engineers wet dream, and management didn't really think it was any need for anything this big. But since this was in the middle of the .com boom, the team got the green light. Engineering called it the "Big Fucking Router" or BFR, and marketing called it "Big Fast Router".
The 12000 or the GSR was introduced in 1996(?) it was wildly successful, and generated 1 billion dollars in sales the first year, and went up from there.
As a result, when the engineers introduced their next wet dream, the HFR or "Huge Fucking Router", the argument was "We can build it faster and bigger than anyone will need, and by the time it is introduced it will hit the market window perfect, and with great success"
The HFR, or CRS-1 is a 100Tbps router. (500 developers for 4 years or $500M).
Only problem: the boom is over, and few are buying.
Solution: Create doomsday scenario that only the HFR can cure.
Just some multiplication: A Youtube stream is 100kbps, so the HFR can handle a billion of these. That is more than there are internet users in the world.
don't cut it off www.mgmbill.org