Interview: AT&T Labs Hossein Eslambolchi

With David Nagel leaving AT&T Labs to run Palm's Platform Solution Business, Hossein Eslambolchi takes the reins of the original networking brain trust. He spoke to Interactive Week's Max Smetannikov.

With David Nagel leaving AT&T Labs to run Palms Platform Solution Business, Hossein Eslambolchi takes the reins of the original networking brain trust. Formerly a senior vice president of packet and optical networks, Eslambolchi first joined AT&T Bell Labs in 1985, studying metro network applications and ultra-reliable communication systems. Eslambolchi, 43, has a doctorate in Electrical Engineering from University of California, San Diego and hold 87 patents. He spoke to Senior Writer Max Smetannikov.

In retrospect, was breaking up AT&T Bell Labs between Lucent and AT&T a good decision?

Yes. The focus was different then. At the time, a certain percentage of the focus was allocated to the service side, and a certain percentage was allocated to the equipment side. It was hard to balance between the two. So it was the right decision.

What are the top priorities for AT&T Labs?

Continue to innovate, bring value-added services, help extract value from existing intellectual property, help develop existing technology, shorten the cycle time from innovation to implementation, bring closer the linkage between technology and operations groups. Also, to innovate in a new level of services. Text-to-speech is one example that we announced a few weeks ago. We have a lot of innovation in the video compression and speech compression technologies. The level of innovation in the field of speech is going by a factor of 20 to 30 higher than the level of innovation in the field of optics. In electronics, there is a Moores Law that says that price performance improves by a factor of two over every 18 months. We are looking at a factor of 100 improvement for photonics every 10 years, and for speech technology its a factor of 250 every five years. There is a lot of potential that you can get out of speech technology that you could apply to call centers. Instead of having a lot of manual intervention and processes, you could have systems that are intelligent enough to have rules-based, policy-based management that would help you deliver new services at much lower cost structure.

How important do you think Multiprotocol Label Switching (MPLS) is, and how well founded are some of your researchers fears that some of its implementations would wreck the Internet?

I believe MPLS is a critical attribute of delivering a lot of value-added services. MPLS is broken down into four categories: You have services that MPLS could support, quality-of-service type plays like voice-over-IP; MPLS could support VPNs [virtual private networks]; you could provide restoration, critical in a mesh architecture; and there is also traffic management. I dont think anybody would debate whether MPLS is a complex protocol or not. In AT&T, we have been in the technology and service insertion business for 100 years. We have been able to scale technologies like nobody else, so we have the core competency to scale MPLS by building revised and sophisticated management capabilities that take the complexity out of MPLS that exists today.

Should public IP networks be more intelligent at the core?

I believe the value of the IP network is not at the core of the network alone. I think the core network is a sufficient condition, but you also need to have access to regional networks so that you have customers on that network. The value of the network is proportional to a square number of your endpoints. The more endpoints, the more eyeballs you have on the network, the bigger the value. Its like a law of gravity - the biggest planet wins. In this scenario, if you have more eyeballs on your network, it gets bigger and bigger. We intend to be the biggest planet.

Should we create an intelligence layer at the core?

I dont think it is practical to assume that we will make the core of the network dumb, and move the intelligence at the edge and vice versa. I think the intelligence needs to go where it needs to go. Some services require the intelligence to be at the core, some at the edge. If you can manage the network to deliver services at the lowest level of cost and the highest level of reliability - thats what you need to be optimizing.

In telecom, most people fall into two categories these days: Bell Heads and IP gurus. Which camp do you feel closest to?

Given my experience, I consider myself an IP guru along with being an operations guru.

Whats your take on broadband?

There is no doubt in my mind that broadband is getting closer to the customer. I believe broadband is a key strategy. We are looking at some of these technologies, DSL being part of this strategy. In the areas where cable is available, I think cable clearly would be a part of this strategy. In the area where we believe you cant deliver either of the two services, you might have to go to the fixed wireless that AT&T Wireless is working with. There are several other technologies we are working on that I am not at liberty to discuss. They would be a combination of wireline and wireless technologies.

What about convergence - will we ever see all services ride over the same network?

At the physical level, yes, you already have one physical network and multiple logical networks on top. Some level of convergence has already occurred. We have been deploying 16,500 miles of fiber - the newest generation of fiber that can handle not only OC-768 but probably OC-3000 plus, when that technology becomes available. On top of that we will have an optical core mesh with DWDM [Dense Wavelength Division Multiplexing] so its a capability of doing wavelength switching, where you could deliver to a customer a wavelength OC-48 or an OC-192. On top of that you have IP, the third layer. Within IP, we will have MPLS. On top of that we will have your hosting centers. And, on top of that, you would have your value-added services, i.e. voice-over-IP, or Web caching, or VPNs.

Are SONET [Synchronous Optical Network] rings going away? I think SONET will stay until the new technology can scale. We are moving from a SONET ring architecture to an optical core mesh.