AT&T VP: 100-Gig by 2010

DALLAS -- Optical Expo 2006 -- Network bandwidth is "exploding" due to demand from applications such as video and multimedia, which will likely require AT&T Inc. (NYSE: T) to upgrade to a 100-Gbit/s backbone by the end of the decade, according to Simon Zelingher, VP of AT&T Labs .

"We will need 100 Gbit/s by the end of the decade," said Zelingher, who was the keynote speaker at Light Reading's Optical Expo, being held here today and tomorrow.

"Currently we are looking at traffic growth that's explosive," said Zelingher. "The past traffic growth was steady and predictable. Now, not only is traffic is explosive, it's unpredictable. All of a sudden, people had a new set of needs. We need to get accustomed to that situation."

Some of the new network needs include the capability to adapt to a wide variety of services, such as VOIP, video, and online gaming, which require low latency and low amounts of packet loss.

"What we see is a lot of multimedia content," said Zelingher. "This includes P2P [peer-to-peer downloading]. It comes in various forms. One is real-time multimedia traffic. This is just beginning to take off. We are going to see a lot more of that."

Zelingher said the rapid adoption of broadband access is pushing the needs of the core network. AT&T has rapidly moved to one global OC-768 (40 Gbit/s) core, but that bandwidth is being eaten up more quickly than expected, he said.

To show how fast the new core, deployed only in the last year, is filling up, Zelingher noted that a 40-Gbit/s DWDM system in one point of presence, capable of carrying 80 wavelengths at full capacity, is already 25 percent full.

Zelingher gave no specifics on AT&T's network equipment, but the company previously announced a contract to buy optical core long-haul gear from Siemens AG (NYSE: SI; Frankfurt: SIE) and switching gear from Ciena Corp. (NYSE: CIEN). (See AT&T Readies 40-Gig Backbone.)

The AT&T network is carrying 5.4 petabytes of traffic every day with 1,600 access points, Zelingher said.

"You can't monkey around with this type of network," he said. "This is the reason why we have to converge the networks. We can't support this kind of scale over multiple networks."

AT&T plans to continue integrating the networks it acquires: first SBC, where integration is still in progress, then BellSouth Corp. (NYSE: BLS) and Cingular Wireless . The last two mergers are expected to be completed later this year.

Zelingher described the single, integrated AT&T backbone as a network that operates in four layers: (1) Hosting centers; (2) IP routers and the "Global Packet Layer," based on MPLS; (3) Intelligent Optical Switching; and (4) Fiber and DWDM. He said this model will be deployed globally.

"Part of the integration is, you have a footprint that is second to none," said Zelingher. "All of this layering has to apply across the entire globe."

Addressing equipment vendors in the audience, Zelingher said service provider optical networking has additional needs, including more standardization that will allow for "plug-and-play" components and better vendor interoperability. He also advised equipment vendors to focus on reliability, especially in software.

— R. Scott Raynovich, Editor in Chief, Light Reading

Belzebutt 12/5/2012 | 3:40:23 AM
re: AT&T VP: 100-Gig by 2010 "Currently we are looking at traffic growth that's explosive," said Zelingher. "The past traffic growth was steady and predictable. Now, not only is traffic is explosive, it's unpredictable.

It was all lies last time, but it's got to be true now! If only Procket and Corvis has hung around just a little longer! Where shall I invest my life savings?
tsat 12/5/2012 | 3:40:21 AM
re: AT&T VP: 100-Gig by 2010 Yeah, we're all gonna be rich!

Seriously, though... does it make more sense for AT+T to pay the huge dollars required for 100gig gear, or does it make sense to equipment companies to keep pushing down the cost of their lower speed (10gig, 40gig) gear?

fiber_r_us 12/5/2012 | 3:40:20 AM
re: AT&T VP: 100-Gig by 2010 The backbone MPLS networks of the major Internet carriers are already using nx10G links for almost every major route in the country. In many cases the backbones are "port bundling" 8x10G links today; and may use more than one port bundle between adjacent backbone cities (ie: 160Gb of capacity between NY and Wash DC via two 8x10G is installed today by at least one of the carriers). The demand for this capacity is still rapidly growing and is somewhere between 50-100% per year (as it has been for over a decade now). Continuing to scale capacity via nx10G or even nx40G is not going to last much longer. So, it is not unreasonable to be considering 100G at this point in time for long-haul networks.

Additionally, interconnects withing the PoPs are also using nx10G interconnects today; so, this is another area that can take advantage of 100G if it were available.
dljvjbsl 12/5/2012 | 3:40:18 AM
re: AT&T VP: 100-Gig by 2010 The backbone MPLS networks of the major Internet carriers are already using nx10G links for almost every major route in the country .... with other examples cited

As a real question, does the market for these types of routes justify, on its own, the R&D expense required for the 100G products?
OldPOTS 12/5/2012 | 3:40:18 AM
re: AT&T VP: 100-Gig by 2010 Didn't the carriers do this to themselves? (See Optical Needs article) So will they pay for it?

Bigger Question of several years ago;

Are they making any money on this backbone expansions??

tsat 12/5/2012 | 3:40:17 AM
re: AT&T VP: 100-Gig by 2010
Right, the carriers recent mantra is "We need cheaper gear"!

And now they are saying "We need you to push the core technology to the bleeding edge"

Which will it be, carriers?
yarn 12/5/2012 | 3:40:15 AM
re: AT&T VP: 100-Gig by 2010 I wonder what is declining faster: the cost per bit or revenue per bit.
Drew Lanza 12/5/2012 | 3:40:14 AM
re: AT&T VP: 100-Gig by 2010 I believe what Zelingher is saying about the backbone and the need for 100Gbps.

There was a great session a couple of weeks ago hosted by OIDA that addresses the technology and market issues associated with 100GbE.

It's worth remembering (and it came up frequently at that session) that 10GbE and 100GbE have fundamentally different market 'physics' than 10MbE, 100MbE and 1GbE.

Earlier versions of Ethernet had the massive desktop PC market to 'pull' them through to volume. That volume, in turn, generated the profits and demand to justify the next generation of Ethernet.

10GbE and 100GbE will not be going to mainstream PC's any time soon. Undersea, backbone, metro, and enterprise aggregation? You bet. But not my PC.

Also, all previous versions of Ethernet were less expensive and consumed less power when used in enterprise applications (e.g. 100m) if they were electrons over copper as opposed to photons over fiber.

The work on 10GbE over copper has shown that at those bitrates (and beyond), optical is both cheaper and consumes less power (i.e. enables higher port density) than electrical.

No one knows if the 100GbE standard will be a 'true' 100Gbps TDM stream or 10 colors of 10Gbps packaged together. It doesn't matter. The break between 'old' Ethernet and 'new' Ethernet took place when we went from 1GbE to 10GbE.

Sign In