Cerf Weighs In on AT&T-vs.-UUNet

The war of words and research building between WorldCom Inc. (OTC: WCOEQ) and AT&T Corp. (NYSE: T) just got hotter.

It started when WorldCom filed for bankruptcy last month, launching AT&T into the catbird's seat while president and CEO-designate David Dorman jawed away about his beleaguered competitor (see AT&T's Dorman Weighs In on WorldCom). Since then, it seems AT&T hasn’t passed up a single opportunity to tell the world that it's ready to take over should WorldCom's UUNet subsidiary go dark (see Whither WorldCom's Network? and AT&T Moves In for the Kill).

Contrary to reports that UUNet carries between 30 and 50 percent of all Internet traffic, AT&T has said that both it and WorldCom carry approximately 13 percent of all traffic each, and that migrating a large chunk of stranded UUNet customers to its network would be no problem at all -- in fact, it would be a pleasure.

Now WorldCom's talking back. In an interview with Light Reading Thursday, Vinton Cerf, WorldCom’s senior vice president of Internet architecture and technology, spoke out about AT&T’s claims.

“They’re suffering from network envy and wishful thinking,” Cerf said. “I think AT&T has a real problem.”

The RHK Inc. data AT&T used to support its network claims is dubious, Cerf says, since no one really knows how much traffic any one network carries (see RHK: Internet Growth OK).

WorldCom, for instance, collects data on how much peak capacity its customers’ networks are utilizing, he says. But that doesn’t translate into actual traffic numbers. “The data are not necessarily commensurate,” he says. “I don’t think it’s accurate.... I don’t see how they could come to the conclusion that we carry [13 percent] each. There’s no data to support that.”

"It's very hard to get traffic data," John Ryan, the chief analyst at RHK, acknowledges. "[But] we've worked long and hard to get it, and we got it." He says his firm used a number of different methods to obtain the numbers it published, including monitoring traffic on several large service provider networks. He wouldn't comment on whether or not RHK had access to monitor UUNet's numbers.

"If WorldCom is claiming that they have no way to measure their traffic... their network is not optimized," says Hossein Eslambolchi, CTO of AT&T and president of AT&T Labs. He says AT&T's traffic measurements are in line with the RHK numbers.

Cerf points to other studies, such as recent reports from TeleGeography Inc., Telcordia Technologies Inc., and Cooperative Association for Internet Data Analysis (CAIDA), that don’t talk about traffic numbers but do show that WorldCom connects to twice as many autonomous systems as AT&T, or that it has many more reachable IP addresses than do its competitors.

Eslambolchi says that WorldCom is missing the point. "Just because you're connected to more networks, doesn't mean anything," he says. "The fact is that we are tied with UUNet [on traffic], and that we have more broadband subscribers." Eslabolchi says that broadband users account for 14 times more traffic than narrowband users.

Regardless of how much traffic each network runs, Cerf says there’s no way AT&T could seamlessly integrate a large portion of UUNet’s customers if the unthinkable were to happen.

“The company suffers from what one could call hubris,” Cerf says. “My conclusion is that the AT&T guys are overreaching if they think they could absorb UUNet’s traffic. Moreover, the likelihood of them having to do that is close to zero.”

That’s because WorldCom has no intention of shutting down its UUNet network, Cerf maintains. “In fact, we’re doing better now than before we filed for Chapter 11,” he says, pointing out that the company is now exempt from paying interest on its debt payments. AT&T might be hoping UUNet folds, he says, “but it isn’t going to happen.”

Furthermore, he says, AT&T’s attempt to lure customers to change providers isn’t going to work. “In my experience, customers don’t move unless they have a reason,” he says. “It’s a big pain to switch service providers. This is the largest broadband network in the world,” he says. “Why would [customers] move?”

What could force a move would be a decline in the quality of service on the network, he says. But he insists that hasn’t happened to the UUNet network.

Up to now, AT&T's downplayed WorldCom’s importance in keeping the Internet up and running. Indeed, several AT&T staff members have been among those alleging that WorldCom executives pumped up Internet traffic growth numbers during the boom years and are at least partially responsible for blowing the Internet bubble that led to the current downturn (see Did WorldCom Puff Up the Internet Too?).

Cerf, predictably, disagrees. “Those numbers were a legitimate characterization of customer demand,” he insists. “If you don’t build the network out to meet anticipated demand, you’re doing everyone a disservice.” He says that the anticipated demand numbers didn’t lie, since during the boom years customers were asking for a lot of capacity.

“If a customer comes and asks for a DS3, your first reaction isn’t going to be, let’s take a look at your business plan to see if this is a good idea,” he says. But he acknowledges that "some people could confuse that for traffic statistics… That’s how some people interpreted it.”

Meanwhile, Cerf says WorldCom is determined not to let its competitors get too far ahead while it struggles through its bankruptcy proceedings.

He insists, for instance, that the company remains focused on developing its next-generation product line. “It’s fair to say that we’re not pouring a ton of money into [new service developments],” he says. “But we’re still continuing with R&D. When we emerge from Chapter 11… we don’t want to be offering two-year-old products.” Last week, WorldCom announced a new global managed firewall solution.

Cerf says he doesn’t know which specific services will drive the industry recovery everyone is waiting for. “Instead of asking what the next killer app will be, you should ask what’s the next killer app that will generate revenue,” he says.

Applications like email and instant messaging aren't going to cut it, he says. One area he thinks might be hot is the entertainment industry, where people seem willing to pay for emergent bandwidth-intensive services.

Cerf isn't expecting to see any signs of industry recovery, though, until the latter part of next year at the earliest.

— Eugénie Larson, Reporter, Light Reading
Page 1 / 2   >   >>
opticalfodder 12/4/2012 | 9:51:01 PM
re: Cerf Weighs In on AT&T-vs.-UUNet 1. Disputed Numbers, but no counter-estimate offered. Nice Spin Move but, ....

2. What is the point in highlighting that UUNET is doing much better since it doesn't have to service debt anymore? Wouldn't a host of companies do a lot better if they didn't have to service debt?

Let UUNET go. Don't argue over traffic - argue over CUSTOMERS. (You two do remember what CUSTOMERS are, don't you? Annoying folks who pay the bills and demand service...)

Let AT&T and Sprint absorb the CUSTOMERS.

If they've got capacity - GREAT.

If they don't have extra capacity - EVEN BETTER. A little DEMAND in the system isn't a bad thing.

I'll bet AT&T and Sprint could craft a plan to minimize TOTAL switching costs for UUNET customers (If auto makers can go 0%, ...) .

Inefficient companies should be squeezed out by the market. Inefficient companies attempting to deceive the market? Throw them in Coliseum and send in the Lions.
puddnhead_wilson 12/4/2012 | 9:50:58 PM
re: Cerf Weighs In on AT&T-vs.-UUNet I think anyone who has reasonably extensive knowledge of how IP, etc., works would immediately undertsnd why it would be impossible to come up with concrete numbers on how much of the total traffic load any one company's part of the network carries.
iptwister 12/4/2012 | 9:50:54 PM
re: Cerf Weighs In on AT&T-vs.-UUNet A previous comment was made that it anyone knowing about IP would know that it is impossible to determine certain traffic numbers.

I disagree.

A reasonable network will carry temporal measurements (usually 5 minute byte deltas) of each interface in the network.

These interfaces can be categorized into 3 (or more) categories:

* 1 * peering (off net / non customer)

* 2 * customer (off net / customer)

* 3 * backbone (inter-router)

It is true that it is difficult to quantify traffic on a flow-basis, but certainly not impossible. Probably impossible given today's actual configured deployment.

However, comparing the numbers for 1 and 2 do give a reasonable measure of traffic.

And therefore it is not impossible.

TT 12/4/2012 | 9:50:53 PM
re: Cerf Weighs In on AT&T-vs.-UUNet GÇ£In fact, weGÇÖre doing better now than before we filed for Chapter 11,GÇ¥ he says, pointing out that the company is now exempt from paying interest on its debt payments.

Thanks for highlighting the anti-competitive nature of bankruptcy. I'm glad you realize how it works - please now consider whether it's fair.

You're better than Worldcom, Mr. Cerf. Don't let Sidgemore and company drag you down with them.
skeptic 12/4/2012 | 9:50:53 PM
re: Cerf Weighs In on AT&T-vs.-UUNet
Q. Who is the director of Network Traffic and
Revenue Analysis at RHK. And where did he
spend the previous couple of decades before
joining RHK.

puddnhead_wilson 12/4/2012 | 9:50:52 PM
re: Cerf Weighs In on AT&T-vs.-UUNet Thanks for the comments on IP traffic, iptwister. Regarding measuring traffic: OK, I agree it might be possible to track raw byte throughput at each node, and use the aggregation of data to create an approximation of what share of traffic a network carries. But a raw byte count is inexact (packets might be better, but even then some message travel through more nodes than others, does that really signify "more traffic?"), and furthermore I don't know if anyone -- or, rather, everyone -- is actively collecting this information, let alone publishing it to allow carriers to compare their raw byte/packet load.

So let me revise my comment to say, IMO an APPROXIMATION of traffic share MAY be possible to compute, but there is no evidence all (or any) of the carriers are collecting this data, let along PUBLISHING it, which would have to be happening to allow AT&T to make this claim.
iptwister 12/4/2012 | 9:50:48 PM
re: Cerf Weighs In on AT&T-vs.-UUNet I am afraid that Mr Pudnhead and myself may actually have a civil conversation of intelligence. I fear that is not allowed on LR?

A few comments in reply, spoken with the most genial of voice:

"might be possible to track raw byte throughput"

- it absolutely is, and is done by every Internet ISP of which I know

"use the aggregation of data to create an approximation of what share of traffic "

- here is the intelligent dissonance, which is that a subset aggregation of traffic data absolutely will be (as) accurate (as the collection/storage methods) (!! should be).

However, the missing variable in the equation is indeed the denominator. These ISPs are loathe to publish their figures, so knowing what is the denominator is difficult. The EU has a fair amount of data from back in 1998, but that#s not so useful now is it?

Therefore, I agree that only an approximation of traffic share is possible without entire disclosure.

Now that said, one can get a fairly reasonable amount of discloser by looking at relative traffic levels with a bit of algebra.

Having done this a plethora of times in the past, any way you slice it, UUNET (AS701) is about 1.5 to 3X the nearest competitor in terms of traffic.

ATT is clumped together in a race for third place with Sprint, CWUSA, and a couple others.

As a final bit of information sharing, using packets as a metric for traffic is not very useful, the byte throughput is most important for various reasons which should be obvious to anyone familiar with the basics of IP.

So yes, I agree, noone knows the exact numbers, but everyone knows the approximate numbers.

And everyone knows ATT Marketing is lying just like they did when they said that if you plugged a non-standard phone into the PSTN the whole thing would collapse.

aa 12/4/2012 | 9:50:44 PM
re: Cerf Weighs In on AT&T-vs.-UUNet He may be a visionary and an accomplished person, but seems to have some problems.

1. Where was he when Worldcom was hyping data
traffic. Now when someone says that the
data traffic numbers were lower than reported
by Worldcom, he jumps and says its difficult
to measure traffic exactly!! Hello, where were
you when Worldcom was hyping those very

2. He says its better now after bankruptcy!!
Who would not be better off by having
thier debt loads taken off thier backs?
I can start a business by borrowing a lot
of money and would have a nice time if
the debt could be removed!!

I think he can be above board and not be
dragged into the mud by his friends at Worldcon!
layer3 12/4/2012 | 9:50:41 PM
re: Cerf Weighs In on AT&T-vs.-UUNet Who really cares ? The question Worldcom and AT&T are both avoiding is why neither can create and maintain a profitable business out of carrying data on their networks. Show me a profitable IP network, THEN I'll be impressed...

broadbandboy 12/4/2012 | 9:50:40 PM
re: Cerf Weighs In on AT&T-vs.-UUNet I think AT&T should be a little more circumspect when they toot their own horn. Their IP backbone network experienced a 2-hour outage the other day in the Chicago area. Router misconfiguration was apparently the culprit.

Their claim to 5x9s reliability just went down the drain!

Anyone have any more details on what happened and why?


Page 1 / 2   >   >>
Sign In