AT&T’s Eslambolchi talks IP

Hossein Eslambolchi is in the unusual position of holding three executive roles at one of the largest U.S. companies. A 17-year AT&T veteran, he is president of AT&T Labs, CTO and CIO. He recently spoke with Network World about his company’s IP plans.

Hossein Eslambolchi is in the unusual position of holding three executive roles at one of the largest U.S. companies. A 17-year  AT&T veteran, he is president of AT&T Labs, CTO and CIO. Eslambolchi says the hands-on experience gives him the diversity of knowledge needed to look ahead to next-generation technology and the operational know-how to reduce costs. He recently spoke with Network World Senior Editor Denise Pappalardo.

MCI has maintained for years that it has the largest IP backbone. But AT&T makes the same claims. Can you explain?We have the largest IP network in the world. AT&T transports about 1,200 terabytes of data per day on our IP network. That’s 1.2 petabytes per day. We have more than 5,000 points of presence. Each day we transport 10 times as much data as voice traffic. We transport about 4.4 petabytes a day of high-speed data including IP, ATM, frame relay and private line. Compared to voice, where we transport 450 terabytes of traffic per day.

MCI keeps saying it’s connected to more autonomous systems or other IP backbones. They also talk about the number of endpoints connected to the network, but those measurements are not sufficient. How much traffic you generate is a more important point.

Why should it be important to a business user if AT&T, MCI or someone else has the largest IP network?

Because the largest IP provider can directly reach more points on the Internet, therefore bringing business users closer to content and also bringing the larger Internet closer to their content. We have 4 million business customers. We’re partnering with six of the largest cable companies to support their IP traffic.

Isn’t quality and reliability a bigger factor for customers?

About three years ago reliability of the network was nowhere near the reliability of the public switched telephone network. We were at about 99% reliability, which is extremely poor. Over the last two to two-and-a-half years, we’ve worked very closely with our vendors. We’ve taken them to AT&T’s school of reliability. When they graduate they’re at 99.99% reliability, which is like a college degree.

What does 99.99% reliability mean to a user?

In determining the reliability of a network you look at the number of defects per million. For example, if a port is available for 1 million hours and it experiences 100 defects in that time, we say it has 100 DPMs. That’s equivalent to 50 minutes of downtime for every 1 million hours.

When will the IP network have the same reliability, 99.999%, as AT&T’s other data services?

It may take another 12 to 24 months to reach that level. Five nines of reliability is what we’re shooting for. A network with 99.999% reliability means it experiences 10 DPM, which is equivalent to a maximum of 5 minutes of downtime every year. At five nines the network is ready for mission-critical applications.

AT&T is consolidating its networks to one IP backbone. What are some of the key steps?

The plan is to deploy multiservice switches, which will support Layer 2 and Layer 3 services, in the network in the first quarter. We’ll deploy 80 switches over a period of time.

…There is a lot of complexity at the edge of the network, and this is where the battleground is. We’ve been building the systems to support this multiservice edge. We’ve also been working on creating an aggregator box called a multi-service aggregator [MSA]. This device will take all traffic in from the edge and deliver it as IP over Ethernet.

Where would these aggregators be deployed?MPLS network or our intelligent optical network.

We have 6,400 buildings directly connected to our local fiber network. These aggregators could be deployed at all of these sites. The IP to Ethernet MSA will hit the multi-service edge switch, which will then connect directly to our 

How is AT&T aggregating traffic today?Lucent and Cisco. Both take traffic from the edge. The only difference is that each use TDM over SONET. That’s still a little inefficient, but that was the best technology available at the time.

We use multi-service platforms from 

So what’s the benefit of IP over Ethernet?I want to move IP farther into the network to get better statistical multiplexing. It will require few service-specific devices, which will reduce the management complexity of the network and also reduce capital expenses.This platform lets AT&T ditch frame relay and ATM-specific gear?Yes. The point I’m trying to make is that a T-1 is a T-1 is a T-1. The only thing that differentiates a T-1 is the protocol that runs over it. We want to move the industry away from controlling these services with different boxes to controlling them with software. This will allow you to change a frame relay T-1 to IP by using a different software module.Are there MSA products out now that meet AT&T’s needs?

We’re in the request-for-information and request-for-proposal process. We’re narrowing down the number of vendors to about five. And then we’ll narrow it down to two vendors and start testing in the first quarter. Everything becomes packet end-to-end, eventually eliminating TDM.

If TDM isn’t used to transport traffic, what will be used?

Next-generation fiber optics is moving toward a photonic infrastructure, which will eliminate the need for electronic cross-connects. Instead, it uses mirrors to direct traffic.

A lot of companies tried this in the past and it didn’t work. But the new technology coming up over the next three to five years will allow you to move from an electronic to a photonic infrastructure. We’ve also been working on long-haul and ultra-long-haul technologies that allow us to manage wavelengths between two cities without the need for regeneration.

You’ve talked about the need to eliminate AT&T’s dependency on incumbent local exchange carriers. What are the biggest problems there?

Access is where the highest level of cost is for us. AT&T spends billions of dollars with the ILECs, which is high-octane profit for these guys. The second problem is the local connection is still a bandwidth bottleneck. Third, there is little way to differentiate your network at the local level.

What is AT&T doing to solve those problems?free space optics technologies. For many businesses fiber is the only answer, but we’re checking out all methods to drive down local access costs.

We’re trying to expand our footprint using local infrastructure. To bypass access today we get [unbundled network element-platfrom] from the ILECs. But we’re looking at all types of technologies that will allow us to bypass the ILECs all together. We’re checking out power line, 802.11, fixed wireless and