Home & Office

Net neutrality extremists should stop playing engineers

This is a lively debate over Net neutrality where Net neutrality extremists prove that they don't have a clue about QoS traffic engineering and don't even understand the very proposal they're supporting. A quick lesson on traffic engineering fundamentals is included.
Written by George Ou, Contributor on

Our own VoIP blogger Russell Shaw has decided to rebut my position on Net neutrality.  Here is my response to Russell.

Russell Shaw says:
George then adds that if everyone is contending for the same bandwidth on an Internet backbone at the same moment in time, then the priority-service packets should never exceed half the total available bandwidth.  He feels no network provider that uses their senses would violate such a policy, because that would cause service to degrade to the point that many customers would be alienated.
Here's where I differ.  I don't think the designation of priority packets should be permitted.  That opens up at least the possibility of favored Internet access to content partners.  Plus, because of hubris, as well as the law of unintended consequences, I think that network providers may overestimate their ability to maintain system-wide quality for those packets that have not been blessed with what I could colloquially call "most favored packets" standards.

Russell, of course you don't think this is a good idea when you have no understanding of traffic engineering.  One of the comments made to me in my blog was that instead of implementing QoS, a Telco should simply add bandwidth to solve the problem.  The knee jerk reaction to this would be "this sounds great" but the problem with this line of thought is that you can add 10 times more bandwidth and you'll still need QoS.
This is because network applications don't behave in a smooth and consistent way and they tend to burst.  It's precisely those bursts that you have to worry about even on a perfectly sufficient pipe that's more than enough to carry all traffic.  So say for example you have a web applications and it bursts really quickly, that's enough to ruin a VoIP call.  Now if you prioritize that Voice packet, all it does is rearrange the Voice packets in a more predictable manner Prioritizing a tiny Voice packet makes someone's call perfectly smooth and clear, but it doesn't actually slow anyone else down at any noticeable levels.
In fact, even congressman Markey's proposal doesn't completely outlaw QoS, what he wants to do is say if you implement QoS for one person then you have to implement it for everyone which ensures that QoS is NEVER implemented because no one will ever offer a free service.  The problem Russell is that you and the MoveOn.org cause is so utterly and hopelessly confused about the subject that you don't even understand the amendment you're supporting.  All you know is that if it sticks it to the Telcos, it must be a good thing.  Otherwise you would have not made the comment "I don't think the designation of priority packets should be permitted".  You say you're for Markey yet you don't even realize that Markey permits prioritization of packets.  If we put people like you in charge of the Internet or any net, you're going to stand there and tell a router engineer to turn off a critical aspect of traffic engineering and that REALLY scares me Russell.  With all due respect and speaking as a former network engineer, you and the other Net neutrality extremists are NOT network traffic engineers and you should stop trying to play one because it is foolish and dangerous.
I'll say this loud and clear; QoS is a reordering of packets that is an essential part of network traffic engineering.  Take the following example where A represents VoIP packets and b represents webpage packets.
No enhanced QoS policy
With enhanced QoS policy
Now note that there are only 5 A packets in the entire stream for either scenario and you still get the exact same throughput for the b packets with or without prioritization for the VoIP A packets.  The difference is that the A packets are now a lot more uniform which makes sound quality go up and the webpage b packets don't really care about uniformity since all they care is that they get there at all intact.  With this QoS example, you can improve VoIP without affecting the average throughput of web surfing.  More precisely, QoS has ZERO throughput effect on non-prioritized when there is zero congestion on the pipe.  If it had been a congested network, then QoS will have minimal effect on non-prioritized traffic.  In either case, QoS is critical and it was invented for a good reason.  QoS is not some evil conspiracy as you seem to be making it out to be.  Your clever designation of "most favored packets" doesn't make a dent on anyone else, but your assault on QoS sure made a dent on a critical aspect of traffic engineering.  I didn't mean to go in to a basic lesson on traffic engineering, but I don't think you can have a valuable discussion on the Markey amendment to Net neutrality without it.  [Update: This is just a simplified representation, there are usually way more b type packets which are 7 to 20 times larger than the A type packets to begin with so it's not like VoIP is hogging any bandwidth.]

Russell Shaw says:
George also finds major fault with the now-defeated Markey bill, which would have banned surcharges for premium service.  I have no problem when a network provider raises monthly subscription fees in markets where speeds have been boosted, but I don't think there should be surcharges for different speeds.

Again Russell, you better think about that in light of the lesson on traffic engineering.  Remember that QoS is not about lowering non-priority traffic speeds at all.  But when you make a statement such as "I don't think there should be surcharges for different speeds", can you hear yourself talking there? Are you suggesting that 6 mbps DSL should cost the same as 1.5 mbps DSL? Hey that's a "surcharge on different speeds" right? Since that's the case, what kind of idiot in the world would ask for 1.5 mbps DSL service if 6 mbps DSL service costs the same because some bonehead congressman thought that would be a good idea? Furthermore, which boneheaded ISP would actually offer 6 mbps DSL service if there can't be a "surcharge for different speeds"? What do you think the end results would be Russell? Think about that for a moment and then answer me if you still think "no surcharges for different speeds" is such a good idea.
Oh and by the way Russell, while you're worried about the nonexistent bandwidth hogging of QoS, you do know that the guy down the street downloading adult content videos even without QoS is putting about 100 times the traffic load as a priority VoIP user right? Furthermore, the number one killer of DSL or Cable performance is oversubscription.  This is where you might have 1000 customers with 6-mbps service but there is only a single 45 mbps uplink to the Internet.  Oh and then there is a sale on DSL service and then the ISP adds another 500 6-mbps customers on that same uplink and then what do you have.  This is precisely why I would love to be able to force all ISPs to disclose their precise oversubscription metrics and other important information.  That alone would do 100 times more than the Markey amendment can ever dream of.

Russell Shaw says:
George then adds: 'Furthermore, if a Telco builds additional infrastructure on top of what they already have for the purpose of transporting Internet traffic, they should not be permitted to designate that entire new infrastructure for priority service and must reserve at least half of that new resource for general purpose 'best effort' service.'
Once again, I disagree.  I find the notion of quotas objectionable.  If all packets were treated equally, true, that might be a form of socialism, but I don't trust carrier-content provider dealmaker cabals.

Once again, you're not understanding what I'm saying or trying to do.  I'm proposing exactly what the true spirit of Net neutrality should be.  Instead of having the ISP adding infrastructure to only suit their highest paying customers, I would mandate at least half of that new infrastructure is designated to general purpose "best effort" class Internet to benefit everyone.  If you're suggesting that any new infrastructure should be completely given to the general purpose Internet, then you are naive because you're never going to get any new infrastructure without some financial incentives for the people investing in the new infrastructure.

Russell Shaw says:
So I am guessing that George and I have a reasonable disagreement driven as much by political philosophy as anything else.

No Russell, I cannot agree that this is a reasonable disagreement.  A reasonable disagreement requires that both parties understands the issues at hand but somehow come to different conclusions.  You and the other Markey-brand Net neutrality backers like MoveOn.org don't even understand what you're supporting and you have proven this to me in your writings and you have no idea of how networking works.  Now there's nothing wrong with an ignorance of network traffic engineering, it's only wrong and dangerous when you and congress try to play the role of a network engineer.

Editorial standards