George Ou, the former Technical Director of ZDNet, has found a new job where he continues to lead the technology sector by publishing innovative thoughts and ideas – sometimes not necessarily his own. Today he is writing to the FCC about Comcast’s experimental “Protocol Agnostic” congestion management method.
George Ou: “The new system will attempt to fairly distribute bandwidth amongst users instead of amongst protocols so that it can be completely accurate and fair.”
So imagine my lack of surprise when I recognize my own concerns echoed in his latest filing
George Ou: “Imagine a user who simultaneously uses P2P and VoIP and they’ve manually reserved about 100 Kbps to their VoIP application by capping their P2P speeds to 2900 Kbps out of a 3000 Kbps connection. But during rare congested times, the ISP can only fairly allocate 2000 Kbps of bandwidth to that users [sic]. Does that mean the P2P application should get throttled to 1933 Kbps while the VoIP application gets throttled to 67 Kbps? If we wrongly assumed that protocol agnostic means all applications get equally throttled, we would be making a grave error since throttling a VoIP stream to 67 Kbps when it needs a minimum of 87 Kbps would simply break the VoIP application completely and the phone call is effectively blocked.”
This is a tortured explanation of my very concern! I mentioned it in forum messages. I described it to the IETF last month in a message. I said, “In particular, I’m concerned with possible scenarios such as someone who is mid-upload of 1100 European vacation pictures (all at 8 Megapixel) when her husband has a medical emergency. She dials 9-1-1 on Vonage or whatever non-Comcast VOIP device. Comcast delays and drops her packets as the new system places her VOIP call behind long packet queues created by other activity in her neighborhood.”
George Ou’s technical solution for Comcast’s Protocol Agnostic bugginess? Don’t be protocol agnostic!
George Ou: “Fairness however does not mean that all protocols should be throttled equally by the network management system, especially when it’s intra-user throttling or throttling between applications being used by the same broadband account.”
So, according to George, Comcast is going to detect VOIP and prioritize it over the other traffic. Cool, huh? No, because in order to do this, Comcast has to look at the contents of all of your packets using Deep Packet Inspection to guess as to whether they’re P2P (bad), Comcast Digital Voice (really, really good), Hulu (competition), Vonage (competition), or DSLReports.com (trouble-making). It doesn’t always work, as Lotus-Notes users found.
Don’t condemn George for this lapse of logic, because it — like a few of his other ideas — really belongs to Richard Bennett, who said to NetworkWorld, “Even in this case where the FCC has banned the used of application-based discrimination, it’s perfectly reasonable for ISPs to discriminate against applications on behalf of a particular user.” Bennett continued, “Say you’ve got two customers, and one is using VoIP and the other is using BitTorrent. You’re going to need to give VoIP traffic preference over BitTorrent in order to ensure quality of service.”
Scorecards — can’t tell the players without a scorecard!
George has shown that Comcast’s proposed Protocol Agnostic scheme has unacceptable side effects — repeating my disapproval to the IETF only weeks ago. Or, from his point of view, I repeated his findings a few weeks before he made them! Either way, he’s done a service and I welcome the fact that he sees the same peril that I do. We disagree as to the solution.
This is a problem of network congestion and the problem is as old as the Internet itself. We don’t need DPI or fancy-sounding schemes to fix the problem. What we need is network management the old-fashioned way — don’t sell more product than you have to offer and upgrade your network ahead of demand.
The Internet is a world-wide network based on a set of accepted Standards. An ISP who invents its own proprietary schemes threatens to break other standards-following applications like Vonage (and who-knows-what-else?). Comcast needs to Stop the Insanity and return our net to the protocol-agnostic Standards-compatible way it was meant to be.
Thanks for calling this to our attention, George, and keep your eyes on your own paper!
Glad to be in the clique,
Robb Topolski
It’s unfortunate that you
It’s unfortunate that you feel the need to adopt the nasty tone that’s characteristic of others at PK, Robb. These issues of network management and Innovative New Applications are often difficult, and there’s no need to attack people for giving technical solutions that can’t be packaged as easily-digestible sound bites. Just because you’re on the PK payroll doesn’t mean you have to lose your pleasant, aw-shucks-I’m-just-a-barbershop-singer demeanor. You could teach Art and Gigi quite a bit about public image. This would be especially useful for them the next time they want to complain about text message blocking on the Internet, for example. Or about something that actually exists. But I digress.
George Ou, like you, doesn’t have a deep background in protocol design or network architecture. But he does have a great deal of hands-on, practical experience and the ability to learn new concepts very quickly. I rate him very highly in native intelligence and diligence, and find no fault with him asking people with more experience to explain things to him. That’s what smart people do.
The problem the ISPs are dealing with today was caused by the acceptance and popularity of an Innovative New Application, peer-to-peer file sharing. It creates a different mix of traffic than previous applications, and managing this traffic on top of the existing traffic created by more traditional apps, such as VoIP and the Web, has given rise to a number of experimental approaches.
The IETF recognizes that the Internet toolkit is inadequate for managing this Innovative New Application, and has created a study group to explore a standard approach. So it’s a fallacy to assert that existing, standard mechanisms are sufficient. The IETF doesn’t see it that way, and they’re the ones who should know.
While we wait for the IETF to develop its standard solution, we’re in a gray area where the ISPs aren’t precisely sure what they can and can’t do, and what will and won’t work, and what’s cost-effective and what isn’t. Consumers who don’t use Innovative New Applications don’t necessarily want to be stuck with the bill for network upgrades they don’t need, and who can blame them? Granny Sue and Uncle Earl never signed up to pay for you to publish the movies you share with the World Wide Internet, and we have to respect their rights.
I personally believe we’re going to solve the P2P traffic problem (much more upload than we’ve ever seen before, by a factor of thousands) through the IETF’s efforts. But in the meantime, it’s reasonable for an ISP to solve the problem by employing a two-stage system:
Allocate bandwidth equally across user accounts, according to instantaneous load; and then:
Prioritize traffic within each account by native requirements, using DPI or any other classification scheme that’s efficient.
Now that’s just my personal theory, and it’s not necessarily the best way for every ISP to approach the problem, but I offer it here as an example of a provisional system that can be put in place while we’re waiting for the good folks at the IETF to recommend a standard solution.
The scenario reminds me of the transition from WEP to AES in wireless LANs. It took the IEEE 802.11 a long time to figure out the ins and outs of AES encryption, the Wi-Fi Alliance jumped into the breach and gave us the provisional TKIP solution. We used that as WPA until AES was standardized, and then we moved to WPA2.
These things happen, and it would be unfortunate for the FCC to jump up and down and get its face all red because the Innovative New Application went on-line before the Innovative New Application’s Management Solution.
So be patient, and don’t advise the regulator to do anything stupid, as they’re likely to take you up on it, and then your barbershop singing will suffer.
And nobody wants that.
Nice comments, Richard. But
Nice comments, Richard. But with statements like “Topolski’s grasp of protocols is consistent with his grasp of arithmetic” on your own blog, I’m not inclined to see your calls for professionalism as very sincere.
Just about everything else you had to say in your reply comment is wrong, too, and I’m not going to spend time refuting that mess. Suffice it to say that your ideas are so off base that you have to have George Ou file them with the FCC in order to get them read — that has to hurt. It would bother me, doesn’t it bother you?
Aw, shucks.
Robb Topolski
Hillsboro, OR
My goodness, Robb, you’ve
My goodness, Robb, you’ve really internalized the PK/Free Press/Google advocacy culture whole-hog: no more Mister Nice Guy for you.
You refer to a comment I made about your claim that the Jacobson AIMD reduces load “exponentially” on packet loss. The correct term is “arithmetically”, as “multiplicative” is not “exponential” decrease. I would be embarrassed to make such an error in a public forum, but I don’t tend to do that sort of thing. You also erred, as you frequently do, in assuming that the keyhole picture you get from Wireshark traces is a big picture, system view. It’s not. There are 1.3 billion people on the World Wide Internet, and Wireshark only shows you the traffic sent to or by from a single one.
But if you can site a single error - just one - in the comment above, I’d like you simply to mention it. You don’t have to try and rebut or explain, just list it. If you can’t, well, that will speak mountains, won’t it?
Cheers
But if you can site a single
It’s spelled “cite.”
Have a good evening, Richard.
Robb Topolski
Hillsboro, OR
So you can’t back up that
So you can’t back up that bluster with a single fact. Sadly, that’s just what I thought.
Well that’s all for you.
this demonstrates the
this demonstrates the mindset of some far out of touch with competition - and the net neutrality that is designed to mimic its absence - they wrap themselves in the garb of a monopolist protectorate fetish and stoop to sensationalist fearmongering worthy of a trash tv crime show - to decide what’s urgent, fair, equitable, appropriate and anything else convenient to throw in the path of net neutrality as something not only insurmountable - even by the dumb-pipe agnostic protocol they railed on as idiotic but now embrace - but downright dangerous and life-threatening … after all, in addition to congestion, there’s also an individual risk of terrorist attack roughly equivalent to getting hit by lightning, so just to be sure everyone is safe, they’ll have to usurp those “neutral” civil liberties as well …
it’s utter nonsense and insulting to conflate the priority of broadband content with congestion or any other source of interruption or delay in order to justify certain centralized command and control actions employed in abnormal situations to intervene in emergency situations … that’s why they’re called emergencies, and chronic, predictable congestion is not an emergency, whether in a traffic jam or peak period of broadband use
the real problem is the refusal by network providers to unbundle not only content from generic broadband service, but to unbundle service quality as well and allow the customers - not the network providers - decide which content is important enough to place in the most premium category - that least likely to get interrupted on a good faith, best efforts basis, an outcome perfectly consistent with competition - which explains why it’s not even on the radar
(one claimed potential problem is that too much of the premium service category could be bought up and crowded out by non-essential services, say by certain P2P wealthtards who insist on P2P with VOIP quality standards - but this would conflict with the free market crowd’s constant harping that the market will provide for those who pay … so for them it can’t happen … at worst, it would drive the price up but remain uncongested)
the arrogance of those who would exploit real emergencies with absurd isolated comparisons of whether medical records or VOIP or whatever content should supercede whatever other content on a CASE-BY-CASE basis, is exceeded only by the ignorance of such a claim to comprehend such omnipotent awareness of everything being sent and received with some grand scheme of a benevolent dictator overseeing what is to pass through or be delayed - it is EXACTLY a net neutral structure which says instead, that there be a systematic, transparent hierarchy in place in advance, to mimimize such central authority overrides in the first place
and this doesn’t mean first-responders of broadband are crippled by neutral rules … if the ambulance service is out and they end up using taxis to send patients to the hospital, likewise, a network provider can switch to P2P or whatever is available to do whatever is possible and judged necessary in a true emergency, but to conjure up the very congestion they have created in the first place as an emergency danger from which they must intervene at will to decide what’s worthy or not is crossing way over the line - if it’s a real emergency, net neutrality is irrelevant anyway
this nonsense comes from those in part, who are cutting the DSL copper line on the way out as fiber is installed, where DSL has it’s own electric supply should the primary source of electricity fail … so they’re concerned about prioritizing medical records or VOIP while they disable the entire back-up connection which could cause sufficient harm in a real emergency to make the delay of medical records from P2P traffic look like a paper cut … even more harm has resulted from the designed absence of broadband availablity to a significant portion of the population … medical records? … how about the forgone huge benefits of online medical research and information in the context of a broken medical care system for those who never get to the point of having a medical record …
Barry, Dead on! I almost
Barry,
Dead on!
I almost want to call their new vision a Congestion Regime, because their actions and words seem to indicate a commitment to maintaining an environment of constant congestion.
This is instead of the traditional Internet management model we have enjoyed over the past 26 years where headroom availability is reasonably maintained ahead of demand (a few rough growth spurts not withstanding).
Robb Topolski
Hillsboro, OR
Robb, Without getting in
Robb,
Without getting in your senseless personal attacks, here is my response.
Fairness is the ultimate end-game in network management http://www.formortals.com/Home/tabid/36/EntryID/88/Default.aspx
Robb May I suggest the
Robb
May I suggest the best way to respond to things like George's comments is to emphasize the facts? Occasionally, it's appropriate to report conflicts of interest, but rarely does that need to be emphasized. If a substantial part of his salary is paid by carriers, a simple mention of that will let people make up their mind about whether it's important. On the other hand, George in my experience can be easily refuted with facts on the net neutrality issues. The details aren't important now, but I've been on a panel with him where I remember him saying "every network must ..." when I have multiple on the record assurances from both Verizon and AT&T their network needn't and doesn't. I have sat at network consoles seeing that directly. Saying Dave Burstein disagrees with George Ou has whatever credibility I've earned; saying the CEO of SBC has testified before Congress in disagreement, as well as his SVP and two SVPs of Verizon is very effective.I don’t have time right now to review the latency engineering with experts, but if the latency problem he suggests was that serious I would expect it to be noticed more often in Skype, FWD, and similar calls (including mine on Vonage with an older router I believe without prioritization.) So I suspect a little research on his claims of p2p packets causing unacceptable latency in voice calls would make things very clear. Our video book is hitting distributors tomorrow, so I have to work on that. But if you haven’t found facts that appropriately answer this remind me to do some research.
“When the facts are on your side, argue the facts” is I think appropriate on this issue. It’s much more effectively persuasive. When Comcast said they were not degrading traffic and a simple, repeatable experiment proved otherwise, that was very persuasive. Contrary to many people’s belief, most of the people at the FCC are willing to acknowledge facts when they are made clear.
db