Security's moving to the network core
- By Joab Jackson
- Jan 22, 2004
Hossein Eslambolchi, AT&T's tech guru
Hossein Eslambolchi holds multiple jobs at one of the government's largest telecommunications providers, AT&T Corp.
Not only is he both chief technology officer and CIO, but he also is president of AT&T Labs, the vendor's R&D arm.
Wearing all three hats, he's in a good position to predict accurately where technology is headed. Eslambolchi determines the company's strategic technology direction, as well as what research it undertakes. He also advises senior leaders on technology issues generally.
Eslambolchi has bachelor's, master's and doctoral degrees in electrical engineering from the University of California at San Diego. He has published articles is 18 technical journals and holds more than 200 patents and patent-pending applications. The New Jersey Inventors Hall of Fame named him inventor of the year in 2001.
GCN associate editor Joab Jackson interviewed Eslambolchi during the AT&T official's recent visit to Washington.
GCN: Is the telecommunications industry still in a state of meltdown? Do you see any growth?
ESLAMBOLCHI: We see growth in data and IP services. It is very clear that global communications is moving toward IP services.
IP lets us be much more nimble in delivering services. A small feature on the public switched telephone network would have taken about two years to develop because of the complexity and proprietary infrastructure. In an IP infrastructure, we could provision the same services in a matter of months.
The telecom industry made a mess back in the 1990s. When we developed a service, it was primarily based on buying from multiple vendors that all had different types of proprietary technologies. Then, when you had to interoperate these services, every domain would have different protocols.GCN: What is AT&T doing to simplify its own infrastructure?
ESLAMBOLCHI: One of the things we have been trying to do is simplify our business processes, operations and systems.
We're moving from a silo-based architecture to an enterprise architecture. By the end of this year, we will be retiring 130 internal and customer systems and replacing them with Web services.
We had seven regional network operating centers distributed across the country. We collapsed them into one global network operations center.
A lot of this has to do with cultural transformation. We used to have a culture where operations managers made their own decisions. Over the period of a decade, you could end up with 50 ordering systems, 50 provisioning systems and 50 billing systems.
As a CIO and chief technical officer, I define what the architecture looks like. To be very blunt, I have to dictate it. We will listen to everybody at AT&T but not let them vote. That has made us a lot more nimble.GCN: Every agency is grappling with building an enterprise architecture. Does AT&T have one? How is it organized?
ESLAMBOLCHI: Yes, we do. The customers define the requirements for us. We don't start our own processes. We define what a customer expects from AT&T, then we design the business processes behind it.GCN: Some agencies are testing converged networks, but CIOs are still concerned about service quality. How deeply into AT&T networks is IP being deployed?
ESLAMBOLCHI: We are already IP. In the AT&T network, the ratio between packet traffic versus voice traffic is 10:1. Our IP network moves about 1,200T a day'that is 1.2 petabytes per day. It's the largest IP network in the world.
IP is no different from Signaling System 7 [the protocol of public switched telephone networks]. A lot of SS7 is already packet-based.GCN: The Defense Department has laid out its plan to migrate to IP Version 6. But adoption of IPv6 elsewhere in government seems to be moving more slowly. When do you expect IPv6 to take hold?
ESLAMBOLCHI: If you take a look at the industry, there was a lot of hype around IPv6 in the late 1990s. Everybody wanted to jump into IPv6. The main reason for it was that we weren't going to have enough IP addresses.
But new techniques such as network address translation solved the problem. With NAT, you just give one router an IP address from the Internet and [assign] private IP addresses behind it. So there was no business logic to move organizations to IPv6.
I think you will have [IPv4] at least for the next three to five years. But one of my predictions is that IP will eat everything, like a Pac-Man game. As more end points become IP, you will need more IP addresses. Because of that, you need to think about IPv6.
The software infrastructure is being developed. You ask Cisco Systems Inc. [of San Jose, Calif.] or Juniper Networks Inc. [of Sunnyvale, Calif.] if they have IPv6, and they will say they either have it or will have it soon.GCN: What should be done about all the network viruses and worms we've been seeing lately?
ESLAMBOLCHI: There is no global oversight of all the IP networks. There are hundreds of thousands of small Internet service providers. Trying to create a global consortium to manage the Internet would be a monumental task, almost impossible.
What you have to do is take a look at where the networks get connected. You have to protect those areas not only internally but also where you get connected.
What we do is build security protection at the edge of the network. We look at anomalies. We use rules-based pattern recognition. We use a lot of tools to protect the perimeter.
Think about intrusion detection devices. People use them at the edges of the network. One of my predictions is that intelligence will go into the core of the network.
With IP, you need a lot more intelligence in the core of the network than at the edges. How do you control a billion end points? Who knows what IP address in China is bombarding the United States with spam?
Let the end points do anything they want to do. You can't stop hackers. But if you have intelligence in the core, as the viruses and worms hit, you can move them somewhere else.GCN: So it's up to service providers to protect their customers?
ESLAMBOLCHI: Industry has spent up to $1 billion on intrusion detection systems, and they don't give a lot of accuracy. CIOs are paying vendors to give them a false sense of security.
If we put intelligence at the core and have the technology to measure and monitor, we'll be able to understand the dynamics, apply rules-based artificial intelligence and let the network say where the anomalies are.GCN: What are AT&T's major research areas today?
ESLAMBOLCHI: Speech technology is growing at about a factor of 500 every five years. One technology we announced a few weeks ago is a voice phone that does speech recognition, natural-language understanding and intent determination. So when you talk with a computer, the computer knows what your intent is.
Say you have a billing problem. Instead of waiting to talk with a human being, you explain what your problem is, and the computer listens and tries to answer your question. And if you still want to talk to an agent, it can route you to an agent. You don't have to press one or zero, or wait for the prompts. That will speed up the process.
Another area is software and systems. We see a lot of deficiencies in the industry, such as security and reliability. How do you drive better software into the infrastructure? So we're working with universities to implant better software training.
A third area of research is around network security and IP network management. We're building technologies to protect the backplane.
If you look at our seven-year horizon, you can see we will still be evolving the company. We've been known as Ma Bell, but we no longer just carry telephone service. We're a managed services company.