Neutral Net Needs Twice the Bandwidth of Tiered

berberine writes with a link to Ars Technica, straight to an article discussing the differences between a net neutral internet and one that supports tiers of content. As you might imagine, our neutral internet is far more bandwidth-intensive; AT&T estimates it might require as much as twice the bandwidth of a tiered internet. From the article: "Corporate sponsorship of research doesn't automatically invalidate that research; what's needed is a close look at the actual results to determine if they were done correctly. According to David Isenberg, a long-time industry insider and proponent of 'dumb' (neutral) networks, the research itself is fine. In his view, it's simply obvious that a dumb network will require more peak capacity than a managed one. But extending that banal observation to make the claim that running a managed network is cheaper is, to Isenberg, not at all intuitive. For one thing, doubling the peak volume of a network does not mean spending twice as much money as it cost to build the original network."

Comments

And

[Umbral Blot]

And sometimes it is worth pursuing an outcome that is not maximally effecient for other reasons, a fact that people seem to overlook sometimes. So what if the internet is half as fast as it could be; that is an acceptable trade-off for a free and open internet.

Re:And

[morgan_greywolf]

Even then, let's suppose you were willing to accept a tiered Internet. How you tier makes a difference in whether it is maximally efficient for a given application. The reason we used managed networks in a corporate environment is because of corporate priorities -- financial transactions are more important than e-mail, so we segment off financial transactions and then give those transactions that must run over the same network as e-mail a higher priority over the e-mail, for instance.

The question is: How do we decide what traffic is more important on the Internet? Who pays? Who pays more? That's stupid. The benefits of a a free and open Internet far outweigh the inefficiencies of working with a basically unmanaged network. (Not that the Internet actually is completely unmanaged -- that's just not true. ISPs shape traffic on their own networks to improve customer connectivity to mail or webservers within the ISP's own network). The point of the Internet is to have a network where anything is possible. Tier it off and you'll make it about as useful as the television networks.

Re:And

[Saint Fnordius]

To put it differently, unmanaged traffic where the drivers get to decide which road they use themselves is less efficient than a traffic net where a central authority dictates to you which highway you're allowed to take. Of course an unmanaged net needs more throughput capacity overall, but in exchange the traffic doesn't require micromanaging. Part of why highways and trucks beat out rail service is because of that flexibility, of not being at the mercy of the switching stations and schedules.

Or consider an irrigation network with multiple sources and multiple outlets. You could either build all the pipes so that any of them could deliver maximum capacity, or have workers actively controlling the valves to distribute the water across the entire net so that one side doesn't overload. The latter solution doesn't require as robust a pipe, but requires a more complex valve system and somebody controlling it.

Re: Re:And

[rnturn]

The point of the Internet is to have a network where anything is possible.

Heh, heh. I can remember when the phone companies wouldn't allow modems because (it seemed to those of us who used them, anyway) it allowed you to do things that the phone company hadn't thought of. "Sending bits across voice lines? NO! You'll have an expensive leased line installed if you want to do that. And you'll lease equipment from us, too. Or we'll cut off your service!"

Tier it off and you'll make it about as useful as the television networks.

You've hit the nail on the head. That's the model the phone companies are trying to emulate. It explains their ridiculous subscriber plans that include "Content by whoever".

I'm not at all surprised at the difficulty that the phone companies are having with the Internet. They had to be dragged -- kicking and screaming -- into accepting packet switched networks in the first place. My guess is that an entire generation of managers (or two) at these companies need to retire before we'll see anything like a basic understanding of the Internet in these companies' actions.

Re:And

[jack455]

Good post, except the implication you make here: "So what if the internet is half as fast as it could be; that is an acceptable trade-off for a free and open internet."
I even agree with the sentiment, but the study incorrectly implied or stated that doubling the peak capacity would double the costs. Even if that were true, not doubling the peak capacity would _not_ halve the "speed" of the internet. For what it's worth here's a selected quote from Ars quoting Isenberg, commenting on the study.

...
doubling the peak volume of a network does not mean spending twice as much money as it cost to build the original network. "The failure of the authors to extend the conclusions from capacity to raw costs of capacity is deliberately misleading," Isenberg says, "especially when the researchers invoked 'economic viability' and 'cost of capacity' in their introduction to the work." ...
According to Isenberg, the cheapest and best alternative is simply to build out dumb capacity: to "overprovision" by as much as 100 percent.

Re:And

[datapharmer]

That's a horrible analogy - It just doesn't hold up. The trains are actually far more efficient than highways as a system, but since the U.S. spends more on cleaning roadkill of the highways than total funding for Amtrak they don't have enough staff or working trains so things get off schedule which causes the entire system to break. Just because the U.S. can't manage a switched system doesn't mean it is bad... try telling someone in France that the highway is faster and more efficient than the train!

Unmanaged networks are inefficient and pointless. There is no damage in routing things to avoid network congestion, but tiered networks are bad too. A tiered network is like a toll road that has restricted carpool type lanes, but the number of passengers doesn't matter - how much you pay in tolls does.

Re:And

[aaronl]

Trains and highway have different efficiencies, though. A train is an excellent way to move a lot of something a long distance. Highways are excellent for non-linear, lowest time transit, or local distribution. I couldn't take a train to my home, for example. I would need to take a taxi, bus, or personal car to get there from the train station. I couldn't take a train to work, since the time lost getting to the train, getting on the train, getting off the train, and getting from the station to work far exceeds the travel time to just drive.

Going between cities is where trains are the most useful. Moving about inside, or around, a city is where the highways are most needed. Rural areas, and there a lot of them in most every country, still need highways nearly all travel.

The unmanaged system of highways allows for all of the same things as trains, though less efficiently, but also allows for *substantially* more freedom of movement and independence of travel time. The right answer, as it always has been, is to use both.

BTW - it isn't just Amtrak that has problems in the US. Nearly all public transit systems are doing their best to approach complete uselessness. It is still faster and less expensive, for me to own, insure, and operate a car where I live than it is to use public transit. This is in metrowest Massachusetts, for reference. NYC is better, but the subway is still no picnic, and light rail can be hell there, too, but it's still a lot better than driving, usually.

For what it's worth, if rail was the better option in the US, business would use it more. As it turns out, you get more for your money by moving things around with trucks and planes. Transit times are much lower, and you can deal with changes in volume and the need to reroute things much more easily.

Re:And

[morgan_greywolf]

According to studies funded by the Mafiaa, the most common traffic on the internet is torrent style P2P network traffic, and movie downloads.

There, fixed it for you.

Re:And

[Maniac-X]

The net neutrality talk isn't making things cheaper, fiber is just becoming more proliferate and easy to manufacture. Overall bandwidth is increasing due to sheer inevitability. The internet will be ruined if it becomes tiered. It stands to put smaller ISPs out of business, and increase net cost for the end user, thus lining the pockets of the major ISP execs... sorry, but that's just not something most people want to see. Let's not turn the Internet into RIAA/MPAA v2.0.

Show me the bandwidth

[Anonymous Coward]

But we're already paying eight times the cost of neutral net bandwidth, so in what way is this study relevant to the consumer?

Re:Show me the bandwidth

[enrevanche]

It is relevant in that this study can be used by the congressman that have been paid for by AT&T to oppose net neutrality.

It is relevant because it will allow AT&T to make a system for which they can charge vastly more than they do now.

It is relevant because it will allow AT&T to reduce your choice more and more over time and to take bigger and bigger pieces of the internet pie.

It is relevant because it will allow AT&T to force more and more companies to deal directly with them for connectivity if their customers want any access to the AT&Ts customers (or shall we call them victims.)

Wait a second...

[vigmeister]

When I am looking at leasing an internet connection at home, I equate bandwidth with speed and this is a reasonably rational assumption (today).

Analyzing the situation and pluggin in numbers,
Assume that the bandwidth available is fixed. What they're essentially saying is that either all of us can get 50BjBps (Bajillion Bps) regardless of the importance of our packets, or using a pareto distribution, 20% of us will get 80BjBps and 80% will get 20BjBps effectively?

I know these are rough numbers, But damn if I know which one I'd prefer... I think at the end of the day, a clearly defined set of standards for prioritization needs to first be developed by an independent body (ICANN/ISO/IEEE?). Once that is done, we can debate net neutrality. Right now, none of us actually know what is going to be prioritized. If streaming video for doctors performing live surgery is prioritized, I'm OK with that. If companies can buy priority for commercial, then I am kind of opposed to it unless I am guaranteed that these priority purchases will subsidize my connection.

Maybe they can have two levels of internet access: Neutral internet access (~$50 p.m) and Tiered access (~$10 p.m). Then let these levels fight it out. Of course, the implementation is unclear to me as I am not network engineer. To think about it, isn't this tiered in itself?

Cheers!

The new way to spin "net neutrality is bad"

[grev]

Majority of people don't know what net neutrality is, they don't care, and they never will. Now, whenever the issue is brought up in the mainstream news or whatever, big business can talk about how it's half as efficient, in addition to being communist and un-American. I can only imagine how this will turn out.

Net Neutrality

[Tuoqui]

Is not about a dumb internet. It is about an internet that does not discriminate based on entry or exit points and/or the protocol being used except where such discrimination will benefit the overall network performance.

Net Neutrality Positive
VOIP Packets receiving priority (because lag and bandwidth throttling reduce performance of VOIP technologies)
Prioritizing Gaming traffic of popular/well used games (IE. MMOs, FPS over internet, etc...)

Net Neutrality Negative
Throttling Bandwidth on P2P applications (This is the big concern on most ISPs, they admittedly do suck up a lot of bandwidth)
Extorting Money from websites who have not paid large sums of money for faster service (YouTube-wannabes)
Delaying or Denying packets coming from X-Network (because they didn't pay extortion money)

Ways to fix things... Run more Fiber. It should not be as hard as it was before since many of the tunnels and such have been made already.

Bandwidth

[Renraku]

Bandwidth is a funny resource.

Imagine if you had a tree that bore fruit once or twice a day. But if you did not eat the fruit within an hour, it spoiled. There's no point in trying to conserve the fruit unless your demand is higher than the output of the tree.

Its always good to have say, 10% free. Out of ten fruit, leave one so that any surprise visitors might have a quick snack as well.

Of course, the other reason you might try to conserve it is to create artificial demand. Now, half of your crop goes to waste. You sell the other half for very high prices saying that your supply just can't keep up with demand and that you must sell them at a higher price due to the whole free market thing.

Point is, every fruit you don't sell will be useless in an hour. But its better to let a fruit rot than to sell it for a decent price, after all.

This looks like double-speak to me

[Fross]

couldn't this be re-interpreted as saying that if they were to run a tiered network, they would have no problem throttling its bandwidth to 50%, in order to ensure the content *they* prioritise gets through unhindered?

Google Translate:

[fishthegeek]

I ran the text through google translate and this is what happened:

Researchers at AT&T were very concerned that bandwidth would be further commoditized if the government does not act to prevent it. If At&t is required to treat everyone the same, then the consumer is free to choose the services that they want based on something called "quality of service" rather than a more practical method of choosing.... say... oh I don't know... uhm... a method of choosing based on how profitable it is for At&t. Having the consumer choose services based on what benefits At&t is a much more practical and convenient way for the consumer to purchase services over the Internet.

At&t is very concerned about the bewildering number of options that the American consumer has available, and with the best interest of our customers at heart, At&t should assist the consumer by limiting the number of choices immediately.

Spokesmen for At&t quickly said that "We do not want to the consumer to get the full unfettered benefit of the Internet because then we would have to actually add infrastructure to meet demand.

Re:In other words...

[bhmit1]

Allowing traffic through requires more bandwidth than blocking traffic.

More importantly, they basically said that when they are allowed to have a tiered internet, they intend on blocking half of the traffic. If you're a generator of traffic to an isp, and you're not paying the tariff/extortion, guess what half you're in?

What Happened to All the Overcapacity?

[rbegga]

To the Bandwidth Providers:

We keep hearing these arguments from the Telco's and Cable COs about how much more difficult it will be to build and maintain an open Internet because of the bandwidth requirements that imposes. Enlighten us as to why this is now a problem considering the major Telecom bust that occurred a few years back was due to the overcapacity you had built into your networks? Google is going around buying up dark fiber from you guys while you're complaining about lack of infrastructure? Nonsense. I don't believe you guys can't figure out a model to make this work for you and us without getting the government involved.

The problem with free market and the internet.

[Repossessed]

AT&T is the primary company pushing to be allowed to do this. I am a Comcast subscriber. This is my traceroute to google.com.

3 ge-5-4-ur01.saltlakecity.ut.utah.comcast.net (68.87.170.161) 9.116 ms 9.247 ms *
4 te-9-4-ar01.saltlakecity.ut.utah.comcast.net (68.87.170.9) 9.021 ms * 9.210 ms
5 12.116.47.117 (12.116.47.117) 19.295 ms 20.255 ms 19.232 ms
6 tbr1.dvmco.ip.att.net (12.122.86.250) 46.279 ms 46.672 ms 45.820 ms
7 tbr2.sffca.ip.att.net (12.122.12.133) 45.180 ms 45.821 ms 45.441 ms
8 ggr3.sffca.ip.att.net (12.122.82.149) 47.504 ms 47.508 ms 47.932 ms
9 att-gw.sanfran.level3.net (192.205.33.82) 167.304 ms 48.359 ms 45.286 ms
10 vlan69.csw1.SanJose1.Level3.net (4.68.18.62) 57.119 ms 49.613 ms 52.738 ms
I also point out that we already have a tiered network. so many MB/s costs so many dollars. Both for the provider *and* the consumer. AT&T is trying to make companies pay *again*. This shouldn't need more laws. This should be classified as extortion.

That said, I'm wary of net neutrality laws, Because from my understanding, the network is already managed. One of the local ISPs CEO did an interview in Wired, where he talked about how his company was already giving priority, based on what customers demanded and what needed the priority most. (VOIP service for example, gets high priority because disruption there matters more than elsewhere.)

This doesn't mean we shouldn't have net neutrality laws, Just that we need to be very careful about writing them, so that legitimate (non extortion) methods can still be used.*

*Though while we're at it, I wouldn't mind seeing it made illegal for college campuses to restrict how dorm students use their internet. There's really no excuse for cutting off somebodies access to communication (IRC is usually the first thing to get hit with idiotic security policies). And from my experience, dorms not only qualify as a monopoly ISP, but typically a mandatory monopoly as well. (I've even seen colleges, public ones, that require freshman to live in the dorms.)

Multicast

[Kludge]

Speaking of reducing necessary bandwidth, when are these ISPs all going to push multicast for media delivery? Isn't this a no brainer for reducing bandwidth?

Re:So, basically....

[DragonWriter]

With all the patching Automatic Update does, I'm surprised that Microsoft isn't all over a neutral net. They may have to pay a fortune to ISPs.

They can afford to pay a fortune to ISPs, especially if it means competitors (like every Linux distro that is gratis as well as libre) that can't instantly suffers a major disadvantage in pushing updates.

It isn't.

[jd]

The study is flawed from a number of perspectives. First, as you mention, the current costs to consumers vastly exceed the costs to provide this mythical doubling in bandwidth. Secondly, network neutrality (to me) means that everyone gets the same experience and access as everyone else. This means that if you implement user-neutral traffic management (packet dropping schemes to minimize retransmits, throttling streams at congested points, fair service to ensure no connection is stagnant, smarter routing algorithms that avoid segments that are basically dead in the water) and user-neutral methods of improving data distribution (web caches, multicasting, SRM) then you gain virtually all of the benefits AT&T claim for their non-neutral system - and more besides - without harming a single user.

The above remedies would give all of the smoothing at peak times on heavily loaded routers, but in a manner that is entirely equitable and - get this - doesn't actually reduce the service provided to anyone. The peaks that kill the backbones are not particularly long-lived and contain a vast number of unnecessary retransmits, inflating the traffic levels. Schemes already exist that can potentially halve the retransmits and diffuse the load over just enough time that it can be handled. Other schemes already exist that can eliminate unnecessary repeat transmissions from source, massively reducing the load on the most burdened segments.

None of these require that any user be given priority or special privileges. None of these require that neutrality be compromised. Yet none of these require that either services or end-users experience any detectable delays (at worst) - and most of the time, both services and end-users will experience a much faster, smoother Internet.

Of course, you'll never get AT&T to admit that the reason they can't do any better is that they're not only greedy but also technologically incompetent. Nonetheless, that is the reality of the situation. It is also something missing from said "study".

An amazingly small multiplier!

[Tired and Emotional]

Twice the bandwidth is an amazingly small multiplier. Its the sort of growth factor that one can reasonably imagine being accomodated with improvements in technology and build-out over a short period. So one has to ask what was the motivation for coming up with that number.

Seems to me what they are thinking is that all the managed stuff will fit within existing capacity and then the unmanaged stuff requires new capacity. Or, to put it another way, all the available capacity needs to be managed.

So the real statement here is "we need to close down the internet as it exists today so we can repurpose the network in order to generate greater revenues".

Two teirs = two separate pipes

[Riskable]

What the big ISPs want isn't just a two-tiered Internet where some traffic gets priority over another. They want two distinct Internets. One were you have control and another where they have control. They'll probably share the same tier-1 backbones but everything below that will be separated (imagine a router configured to send packets from their sources directly to you via a hyper-speed backbone whereas all other traffic gets routed through a dozen or so more hops on the "economy" backbone).

If you want a practical example of precisely how they they plan to violate network neutrality look at the DOCSIS 3.0 spec. It reserves about 80% of the bandwidth on the coaxial cable for video and telephone services that are exclusively provided by the cable company (i.e. no one else is allowed on). The other 20% of the bandwidth is provided as general Internet access (with the usual limited upload speed). This way they can be the gatekeeper for high-bandwidth content (i.e. video) and low-latency applications (i.e. VoIP) while every other business that wants access to their customers has to either pay to get on their high-speed channels or get stuck with the slow lane.

The telephone companies are already rolling out technologies that divide up fiber connections in a similar fashion. The "big plan" is to get paid extra for that exclusive, high-speed and low-latency channel into people's homes. It is a hugely anti-competitive situation.

If you provide streaming video to anyone on the Internet you will not be able to compete with the speed and quality of the video coming over Comcast's, AT&T's, and Verizon's dedicated pipes. If you're a VoIP provider that provides telephone service to anyone on the Internet you will not be able to compete with the low-latency and high quality of the big ISP's dedicated pipes. If you provide *any* service over the Internet all it will take for you to be crushed out of existence is for the big ISPs to start offering the same service on their dedicated, exclusive channels.

It isn't about prioritizing traffic. It is about dividing it up and destroying the free market that is Internet access in people's homes. It is literally "divide and conquer".