Slashdot Mirror
Net Neutrality vs. Technical Reality
penciling_in writes "CircleID has a post by Richard Bennett, one of the panelists in the recent Innovation forum on open access and net neutrality — where Google announced their upcoming throttling detector. From the article: 'My name is Richard Bennett and I'm a network engineer. I've built networking products for 30 years and contributed to a dozen networking standards, including Ethernet and Wi-Fi. I was one of the witnesses at the FCC hearing at Harvard, and I wrote one of the dueling Op-Ed's on net neutrality that ran in the Mercury News the day of the Stanford hearing. I'm opposed to net neutrality regulations because they foreclose some engineering options that we're going to need for the Internet to become the one true general-purpose network that links all of us to each other, connects all our devices to all our information, and makes the world a better place. Let me explain ...' This article is great insight for anyone for or against net neutrality."
I think the article has some valid points regarding the technical aspects of the Internet, but I don't understand why those aspects make net neutrality legislation a bad thing. My understanding of net neutrality is that people want the Internet to remain neutral. They do not want providers to charge favorable rates to their friends and extortionist rates to their competitors. They do not want small ISPs forced out of the market. They do not want websites and users to be double-charged for the same use. I don't see how any of these issues are technical. I don't see how legislation that would keep things fair also would eliminate an ISP's ability to improve the performance of jitter sensitive applications as well as jitter insensitive applications. I mean you could argue that it'd be legislated wrong, and you'd probably be right. But from a technical standpoint, assuming it's legislated correctly, why is net neutrality technically impossible? Or am I completely misunderstanding the net neutrality issue?
He completely ignores multicast in the paragraph about HTDV being trouble for the Internet, and someone should at least explain why it's not relevant. Otherwise it kind of sinks his battleship w/r/t that argument, IMO.
Multicast only works if internet TV is going to be like regular TV where a show is aired at a particular time. If it's going to be more like youtube on steroids multicast doesn't help.
ccalam - acoustic versions of new songs.
A related point that seems to run through the article is that more bandwidth is not the solution. But he doesn't explain why - for example This problem is not going to be solved simply by adding bandwidth to the network, any more than the problem of slow web page loading was solved that way in the late 90's or the Internet meltdown problem disappeared spontaneously in the 80's. What we need to do is engineer a better interface between P2P and the Internet, such that each can share information with the other to find the best way to copy desired content. In the first case I think he's completely wrong, more bandwidth is exactly what solved the problem. Both in the network and the applications use of that bandwidth (netscape was the first to do simultaneous requests over multiple connections - which did not require any protocol changes). In the second case, he's talking about Bob Metcalf (the nominal inventor of ethernet and nowadays a half-baked pundit) predicting a "gigalapse" of the internet specifically due to a lack of bandwidth...
It's interesting to note that ATT themselves have declared more bandwidth to be the solution. They didn't phrase it quite that way, but ultimately that's the conclusion an educated reader can draw from their research results. 1x the bandwidth of a 'managed network' requires 2x the bandwidth in a 'neutral network' to achieve the same throughputs, etc. Sounds like a lot, but then you realize that bandwidth costs are not linear, nor are management costs. In fact, they tend to operate in reverse economies of scale - bandwidth gets cheaper the more you buy (think of it as complexity O(x+n) due to fixed costs and the simple 1 to 1 nature of links), but management gets more expensive the more you do it because the 1-to-1 nature of links gets subsumed by having to manage the effects of all connections on each other n-to-n style for O(x+n^2). Ars Technica analysis of ATT report
When information is power, privacy is freedom.
"I know that's not true. The Internet has some real problems today, such as address exhaustion, the transition to IPv6, support for mobile devices and popular video content, and the financing of capacity increases. Network neutrality isn't one of them."
The effen telcos already got paid 200 billion dollars to do something about getting fiber to the premises and blew it on anything but that. Where's the "political engineering" solution to look into that to determine where the "QOS" broke down at ISP intergalatic central? Where are the ISP and telco fatcats sitting in front of congressional hearings explaining what happened to all that freekin money? Where did it go, real facts, real names, real figures.
And why in the hell does the bulk of the public air wave spectrum only go to the same billion dollar corporations, year after decade after generation, instead of being turned loose for everyone-you know, that "public" guy- to use and develop on? Why the hell do we even *need* ISPs anymore for that matter? This is the 21 st century, there are tons of alternative ways to move data other than running them through ISP and telco profitable choke points, and all I am seeing is them scheming on how to turn the internet into another bastardized combination of the effen telco "plans" and cable TV "plans". Really, what for?
Where's the meshnetworking using long range free wireless and a robust 100% equal client / server model that we could be using instead of being forced through the middle man of isps and telcos for every damn single packet? And what mastermind thought it was a good idea to let them wiggle into the content business? That's a big part of the so called problem there, they want to be the tubes plus be the tube contents, and triple charge everyone, get paid both ends of the connection and a middle man handling fee for..I don't know, but that is what they are on the record wanting, and industry drools like this doofus are providing their excuses. Not content with hijacking all the physical wired reality, for 100 years now, they get to hijack all the useful wireless spectrum, and no, WIFI DOESN'T CUT IT. That's at the big fat joke level in the spectrum for any distance.
If you want on-demand, and NO local storage, then you are indeed in trouble.
The problem is subtle, and I've only seen it now that I read the TFA although I've experienced it with our internet connection at work.
The sliding window mechanism of sending packets before the ACK of the previous one until you get NACK and then back off has an unpleasant side-effect. An ACK train coming back over three hops from the local P2P clients or ISP-based servers moves faster than one heading across the world over 16 hops with higher ping times. Therefore the sliding window opens more and the traffic over the three hops can dominate the link.
Now add that problem with BitTorrent clients reported earlier that try for max bandwidth. That can force the window even wider.
And once the DSLAM/switch/aggregation port is saturated with traffic, it will delay or drop packets. If those are ACKs from the other side of the world, that window closes up more. There goes the time-sensitive nature of VOIP down the toilet.
On a shared-media network like cable, it doesn't even have to be you. If two people on the same cable are P2P transferring between each other, there goes the neighborhood. They dominate the line and the chap only using Skype down the road wonders why he isn't getting the performance he needs.
I'm opposed to price-oriented non-neutral networks, your ISP charging Google for your high speed access to them. But a non-neutral network that does proper QOS by throttling bandwidth-heavy protocols that don't behave themselves on the network is acceptable. As long as the QOS only moves the throttled protocols down when needed.
Trying to become famous by taking photos. Visit my homepage please.
Just forget about Multicast, it's a dead-end idea. Not because it's technically flawed (actually, it works pretty nicely), but because it ignores economics.
A simplified economic model of the Internet calls for multiple level of service providers that sell bandwidth to each other. So I, as your ISP / backbone provider make as much money as bandwidth you can use. I have the option of enabling a technology that allows you to be more efficient and use less bandwidth, therefore pay me less. Meanwhile, this technology offers no benefits for me, in fact costs me money, the money needed to implement it and manage it.
To add insult to injury, this technology works properly only if all the hops between you and your destination have deployed it correctly. So a bunch of telcos who's primary business is selling bandwidth must go trough hoops to make your data transfer more efficient. No, it's not gonna happen.
To be successful, Multicast must be completely redesigned from an economical perspective such as to provide a immediate benefit for the provider that uses it (if this is at all possible), without reducing his revenue potential.
they should just quit offering unlimited data plans unless they can actually offer unlimited data. unlimited dialup is easy to provide for, as in a whole month a user can get at best theoretically about 12gb if they are continuously downloading at full speed.
the real problem is the marketing people are defining service options that the networks are not capable of supporting. some services are making a profit to support other services that aren't, which is fine in, for example, pre-packaged computer bundles, but because with internet service this affects everyone, this is the end result - isp's don't have the ability to provide the level of service they advertise so they must resort to throttling, which is of course done arbitrarily to certain kinds of traffic as they are the biggest bandwidth users, rather than doing it generally.
if isp's just didn't spend so much time trying to hook those high bandwidth users up and made the prices of service to them higher, then the isp's could spend more money enhancing their bandwidth capacity instead of ending up having to explain why and what traffic they are shaping to keep use within the parameters of their networks.
there is many factors related to how network applications are written, how various tcp/ip stacks schedule, how effective QoS systems are, and how widely deployed they are, but there is one guaranteed way to ensure networks aren't bogged down by bulk traffic and streaming users - always keep traffic levels below about half of capacity. the line might be rated to transport data at a certain speed but when you fill that pipe past a certain point you wind up with a great deal of turbulence which leads to latency issues.
it's a bit similar to mastering levels in audio engineering - sure, you may have 120 dB of resolution in your recording medium, but the closer you get to filling all that space the less headroom you have for periodic spikes, which has lead in the commercial music engineering to more use of dynamic range compression, which produces a much 'duller' sound with less dynamics (some even say that this compressed dynamics leads to fatigue in the listener) - this problem never happened in cinema sound engineering because someone set a standard for how many dB average power should be targetted in a mix. Similarly, if the network provision industry would set a standard of aiming at around 50-60% utilisation average and accordingly adjusted planning for bandwidth upgrades and market penetration none of this would be a problem.
beancounters see the network capacity specification and expect that they can run the network at that level without any problems. But of course beancounters also rate the potential of a resource according to a percentage of customer turnover below a certain level, meaning they can cheapskate to some degree and of course being that businesses care more about the bottom line than good service, this is the sort of issue that cannot be solved by anything other than legislation.
i believe network neutrality as a concept misses the real point at issue here, which is simply businesses squeezing more money out of their lines than it is possible in real practise to allow, and pushing this limit just short of messing up the whole network. throttling bittorrent and streaming video is all about trying to hold back the flood of bandwidth demand so they can put off the upgrades for longer.
there would not be a problem if they just didn't provide more bandwidth on the local loop than can be carried through the peering connections.
I'm opposed to price-oriented non-neutral networks, your ISP charging Google for your high speed access to them. But a non-neutral network that does proper QOS by throttling bandwidth-heavy protocols that don't behave themselves on the network is acceptable. As long as the QOS only moves the throttled protocols down when needed.
Thank You!I work for an ISP, and net neutrality scares the hell out of me. We do not want to, and will not throttle back certain sites who won't pay us for premium access, or create a tiered pricing structure for our customers. What I want, is the right to manage my network to give my customers the best performance by de-prioritizing badly written, and poorly behaving protocols, AKA: 99% of all p2p stuff.
We also don't want to see content providers shift their bandwidth costs onto the ISP networks via the use of p2p. Why pay for expensive backbone links when you can shove 50% or more of your bandwidth onto your customers, and their provider's network? Either let us ISPs manage our networks, or we will start charging for upload bandwidth on a usage basis. I really don't want to do this, but if net neutrality becomes a reality, I see this becoming a very popular way to save on bandwidth costs. Blizzard already does it, patches for World of Warcraft are distributed via bittorrent. Why they think it is appropriate for their service to be offloaded onto my network is beyond me, but they do. When I can't rate limit bittorrent, and it becomes a huge bandwidth hog, my customers that patronize services that are the source of the problem will see their bills go up.
Thank you, I finally read a post from someone who gets it. I didn't think that would ever happen.
Oh, and any replies to the effect of, "well, its your own fault for not having enough bandwidth" can just go eat a dick. I have bandwidth, and that is not the point. The point is content providers should provide their own bandwidth, not leach it from the ISPs in the name of the heavenly, super great, don't ever question it, p2p software demi-god.
Man, I got way off target there.
--Nuintari
slashdot : where an opinion can be wrong.
Here's how media companies will kill the free internet we all know and love:
The result will look like broadcast media does today, one big corporate billboard, instead of a free press. Just a little censorship is like being just a little pregnant.