Noted protocol designer Krishan Sabnani talks about next-generation networks.
Krishan Sabnani is the vice president of the Networking Research Laboratory at Lucent Technologies' Bell Laboratories. In his personal research he has made major contributions to the area of communications protocols, and has designed several protocols such as SNR, RMTP and Airmail. He has also made significant contributions to conformance test generation, protocol validation, automated converter generation, and reverse engineering.
UBIQUITY: What's your day been like so far?
KRISHAN SABNANI: This morning I met with a start-up company that's been focusing on the Enterprise space for 802.11 networks. My business colleagues and I are exploring whether Lucent would benefit from some kind of a business arrangement with them. Then I had a meeting with our friends in our Data Networking business unit on how MPLS can be deployed in different networking products so that they would internetwork and provide differential advantages for our product lines.
UBIQUITY: When did you make the change from researcher to research manager and research vice president?
SABNANI: I was a researcher for about 12 years from 1981 to 1993. I became research manager in '93. I was a director from '93 until 2000, and then I was appointed to my current position of leading the Networking Research Lab in the beginning of 2000.
UBIQUITY: How do you like managing research, as contrasted with doing it?
SABNANI: It has been wonderful. When you are an independent contributor sometimes your influence is limited by the amount of time that you have to personally spend on your projects. I'm now leading a team of outstanding researchers -- many of whom are a lot better than I am. So I have an opportunity to provide direction, support and guidance in a way that has far greater impact on the business. I also derive great satisfaction by playing a role in the growth of these talented people.
UBIQUITY: Tell us about your intellectual development. What did you want to do as a schoolboy?
SABNANI: I wanted to be a scientist in those days, maybe a physicist. Once I completed high school, I entered an institute for technology in India, IIT Delhi. This is how I got into engineering and computer science. When I was an undergraduate I had a lot of interest in computers, and this interest continued to grow.
UBIQUITY: You have been in the educational system in three continents -- an undergraduate degree in India, then in the Netherlands for a master's degree, and then at Columbia for a Ph.D. What thoughts do you have about your different experiences in different educational systems?
SABNANI: The systems here in the US and India are really outstanding. In India, there is a very systematic approach whereby the Indian Institutes of Technology get the best undergraduate students from all over the country. So you get to interact with the very cream of the crop. Here in the US, we get the best researchers and students from all over the world, particularly for graduate programs, and that's even better.
UBIQUITY: Thinking about your management of research, what do you find the most difficult to do? Is it recruiting people or what?
SABNANI: Recruiting top researchers is always a great challenge. I have managed to be quite successful in this area, but it took some time. In the earlier stages when an organization does not have a number of good researchers it's particularly difficult. Once there's a critical mass it becomes a little easier. But the next step after hiring very good people is to make sure they grow into productive researchers.
UBIQUITY: How do you mentor them?
SABNANI: It's always a good idea to keep giving them tough challenges, along with constructive feedback. It's very important to make sure that we are strongly supportive of all of our researchers; but at the same time it's always a good idea to have a look at their experience and see how they can grow and become even more productive. Contrary to some views, it's not a good idea to leave researchers alone to do whatever they want to do. It's important that they are aligned with the priorities of the organization, so that they're a part of. This is one area where mentoring is particularly important.
UBIQUITY: How many people do you have in your group?
SABNANI: About 70 researchers.
UBIQUITY: How often are you able to interact with them personally?
SABNANI: I interact with my directors at least several times a week, and some subset of researchers I would say at least every couple of weeks. But I get through to every member of my lab once every two months.
UBIQUITY: Do you have any general ongoing programs like weekly seminars or something like that where the whole 70 get together?
SABNANI: Yes, we do. We have regular meetings when members of the lab get together to talk about their work. These are highly interactive sessions.
UBIQUITY: How do you think Bell Labs has changed during the time you've been there and from what you know of before then?
SABNANI: I've been with Bell Labs now for more than 20 years. Both the company and the industry around us have changed. When I joined Bell Labs it was a part of a regulated monopoly with more than a million employees. Now it's a part of a much smaller company. In the old days when it was a regulated monopoly research funding was guaranteed. Even in the old days it was important to make sure that the rest of the company felt it was getting value for the money it spent on research, and now that's even more important.
UBIQUITY: And what do you think is the most important way in which the industry itself has changed?
SABNANI: Back in the early '70s and '80s, product cycles used to be five to six years. But now in this new age product cycles have become much shorter. This means that for our researchers to have their ideas become real they have to work a lot more closely with business units to minimize time to market. Otherwise it will be too late for the idea to be viable. You've heard of Internet speed. Many new products come from these pressure-cooker Internet startups. We need to compete with these speeds.
UBIQUITY: Are your researchers all happy about the need to productize quickly -- and, for that matter, the need to productize at all?
SABNANI: As the business has changed so have the attitudes of our researchers. They have always wanted to do what is important. Our researchers take pride in making sure their ideas get widely accepted. So they are very anxious to work with the business units -- not as developers, but in a role where they can add the unique value that they can bring. They want to come up with innovative ideas that can create a competitive advantage for the products or the business processes. And they will do whatever work is necessary, including even writing key programs and working very closely with the development teams.
UBIQUITY: And they feel no particular urgency to do so-called basic research?
SABNANI: Parts of my lab do basic research. Everyone in my lab is working on creating new systems. I'll take the example of two great scientists and engineers -- Einstein and Edison -- to illustrate what the difference is. I would call Einstein a scientist or analyst, and Edison a great inventor. The majority of people in my lab are more like Edison -- they invent and create new systems. But we also have some people who are primarily analysts. The important thing is to excel in what you do -- either in scientific leadership or innovations that have an impact on the business.
UBIQUITY: What are your own general feelings about new technology? What do you find disappointing in networking technology right now?
SABNANI: The most disappointing thing in recent years has been that we have this great Internet and Web technology, but recently it was oversold. The overselling created a bubble, which has burst and has caused catastrophic problems. A lot of promises were made and the promises have not kept up with reality, although I think many of those problems potentially can be solved as we go forward. That has been the most disappointing thing from my perspective.
UBIQUITY: Which promises are you thinking of mainly?
SABNANI: Basically that we will provide these Internet services and they will take off immediately. Many of these new services that were created -- for example, multimedia services on the Internet and even voice over IP -- have not done as well as originally promised. Again, this is not the result of a fundamental problem with the technology. The problems will be solved eventually.
UBIQUITY: A term that peppered every technology conversation a few years ago was "convergence." All kinds of things were going to converge: the boxes were going to converge, technologies were going to converge, companies were going to converge. What do you think of convergence as an idea now?
SABNANI: I think convergence is another term that has been oversold. Some products will converge and some won't; in fact in some cases the trend may be the opposite. But I think that convergence will occur with respect to end-to-end network performance. This will happen increasingly as we find ways of providing the right quality of service and ways to manage these new packet networks. It's already happening to some extent but I think you'll see more as we go forward.
UBIQUITY: What will it look like?
SABNANI: The speeds are going higher and higher. Terabit core routers will be deployed. You will have optical switching cores. The core will become faster and faster, but a lot of the intelligence will migrate to the edge. Edge routers will have a lot of intelligence and provide many new services. This will be the network of the future. The best of the technologies will be combined. The control plane in many of these networks will be MPLS-based.
UBIQUITY: Will the end user feel this convergence as something that happens in a particular box? For example, will television sets and personal computers eventually become essentially the same thing?
SABNANI: This may be a good example of the overselling. I'm doubtful about it because TV sets are connected to a broadcast channel. That's a mature technology that works well. Computers will be used for interactive services; they will also absorb services that have been performed by telephones. For example, five years ago if I wanted to find out when an airplane would arrive at Newark Airport I would call an 800 number. Now I can get on the Internet and track an incoming flight. So the convergence will happen more along those lines, not between TV and computers. It is possible that you might be able to use the same displays for both for television and computers. Computers, as well as telephones, are connected to bi-directional channels. Anything that is bi-directional is much more expensive than a channel that usually sends information only in one direction -- for example a fax line channel or a cable. Many people are talking about providing TV services over these bi-directional channels but I don't think this technology will be able to compete with purely broadcast channels.
UBIQUITY: Speaking of channels, what are your thoughts on the apparent glut of fiber in the world?
SABNANI: All of these highways have been built. We need to build new access ramps onto them to be able to utilize the capacity. For example, here in New Jersey -- and I would say this is true in most of the US -- high-speed data services over cable are really doing well. DSL has not yet taken off but I think it will as we go forward. Once these broadband access networks are widely deployed I think most fiber capacity will be utilized.
UBIQUITY: So it's just a temporary glut?
SABNANI: Yes, it's temporary. In fact, Internet traffic is still growing at a rapid rate. One day we'll get to a point where all of the surplus capacity has been used up.
UBIQUITY: You mentioned DSL. How you feel about DSL versus broadband over cable?
SABNANI: DSL will be provided mostly by RBOCs (Regional Bell Operating Companies) like Verizon Communications. DSL will be wonderful for interactive, high-speed data services. In order for DSL to take off, it needs to do a better job of handling video. But DSL has some potential advantages over cable. For example, it should be able to provide high-quality voice services, which will be difficult to do over cable networks. A lot of people are talking about video over DSL, where you use a compressed video stream from the network to homes. Signals can be transmitted at a couple of megahertz per second using new video and multicasting technology. It's not really clear whether this approach will take off. Again, I think DSL will be great for providing the Internet services, but I think cable is better for sending TV signals. Cable companies have done a great job in providing data services to homes. In fact, they have been a lot more aggressive than RBOCs in developing and marketing this technology. That's why in New Jersey, for example, cable service is doing a lot better. But I think it would be very difficult to put high-quality voice on these cable networks.
UBIQUITY: We started out talking about what you did today. Has it been a fairly typical day for you?
SABNANI: My job is a combination of working with my researchers and interacting with people outside my organization -- e.g. people in our business units and people in other Bell Labs organizations. Typically I spend more time working with my researchers than I did today. We usually work on conceiving and launching research projects and programs and removing barriers to progress.
UBIQUITY: Well, let's go from your workday to your environment. Describe your computing environment in your office and at your home.
SABNANI: I have PCs in my offices in Holmdel and Murray Hill, NJ that are connected to a LAN. I also have Unix service on them. At home I have an ISDN line and a cable connection. I also have an 802.11 network at home.
UBIQUITY: Do you see an ordinary person's information environment changing drastically over the next ten years or so? Would your neighbor likely be doing a lot more of what we call "computing" than he or she is already doing today?
SABNANI: I think many of the things that people do today, say, regular mail and using the telephone, will move more and more onto the network. I think fax machines will disappear. Most of us use Internet for ordering new things, for getting bus, train, and plane schedules, doing our banking, even paying bills. This trend will continue. To do each one of those things you have to run programs on some computers. People will also be doing much more computing on hand-held, wireless devices. But I don't think they will be writing new programs on their computers. That will continue to be done by only a select number of people.
UBIQUITY: Speaking of wireless, what about the new wireless environment? Do you use wireless much?
SABNANI: I use a cell phone and also have an 802.11 network at home. My wife is a physician, so she, of course, has pagers and a cell phone. There's no question that wireless technology is going to change the world. The question is how well 3G data will do as we go forward. Gradually we'll fix problems with that also. If we can integrate cellular service with wireless LANs, those services will take off.
UBIQUITY: Is there a way you could generally characterize the nature of the problems with the 3G wireless right now?
SABNANI: There are two versions of these 3G networks. There are the ones that are being deployed here in North America called CDMA2000/1x. They have been deployed in parts of the United States and are doing well for voice services. But data services still have some problems: long delays and division of functionality between multiple boxes, which reduces performance for data services. As people learn more about where the bottlenecks are they will be fixed. In the US you will soon see new technologies introduced e.g. High Data Rate (HDR). This is a Qualcomm standard that we will be deploying that will provide at least a few megahertz per second in these 3G networks.
UBIQUITY: And in Europe?
SABNANI: UMTS (Universal Mobile Telecommunications System) has many technical problems in deployment; they are still trying to figure out how to make these wideband networks work. Once they have solved those problems they will discover problems with data services similar to the ones we have here. Another problem in European networks is that they do not use pure IP or mobile IP; they have their own standard.
UBIQUITY: What are you particularly proud of that you've accomplished?
SABNANI: One of the personal research contributions I'm most proud of is the invention and implementation in the lab of RMTP (Reliable Multicast Transport Protocol). In my management career one of my most satisfying activities has been working with a group of researchers and our business unit colleagues to create next-generation wireless network designs and architectures, which have had great influence here. One example of our work was the creation of asymmetrical protocols that allow most of the processing to be done in the base stations, which allows the appliance to be very simple.
UBIQUITY: Is the technology vision that Bell Labs has much different from the shared wisdom of the entire networking industry?
SABNANI: The networking industry has multiple segments. There is a telecom view and a networking view. The vision that we have in my lab is that as we go forward in wireless networks, you will see a discontinuity from the telephone networks, and eventually wide deployment of pure Internet protocols. This vision differs from what most people in the cellular and PCS industries think will happen.
UBIQUITY: Who's right?
SABNANI: Only time will tell who's right.
UBIQUITY: Looking back, has anything really surprised you about the way things have developed?
SABNANI: One of the things I found most surprising was the rapid change in mindset throughout the industry caused by the Internet technologies. Suddenly people believed that small companies that adapted these technologies would be able to outdo the established players. There were a number of skeptics, including me, at that time. Most of these skeptics changed their minds because of the meteoric rise in the capitalization of these companies. The ratio of capitalization to revenues became a major figure of merit for a business. Then the fall came. This rise and fall had a big impact on all of us.
UBIQUITY: Are you saying that there were two separate surprises -- that both the rise and then the fall were two separate surprises?
SABNANI: Yes. So many people bought into all the hype so easily. Now the opinion seems to have shifted to the other extreme. Still, we have a great technology here, and it will continue to change the world.
UBIQUITY: Your wife, you said, is a physician. Does she think that your technology is going to change her career, her life, her profession, her world?
SABNANI: Yes, I think so, because everyone has started using the Web. I see her logging onto the Web to get her medical information. Everybody's using e-mail, and most physicians in the old days were using mostly fax machines and pagers, and maybe cell phones, but that is changing. Some of my wife's colleagues even read sophisticated images over the Internet.
UBIQUITY: What's on your own agenda now?
SABNANI: As I indicated earlier, my vision of the network is that the core will become faster and dumber, and the edges will become more intelligent. I want to help my company play a leadership role in making that happen. Also, in wireless we want to solve some of the key problems with 3G networks. We are trying to solve the problems in building these networks with packet backbones. Another area that we are looking at is how these networks will evolve. If you look at the current wireless networks, you see million-dollar base stations, with a T-1 line, or E-1 line, going into a central office. But now there's a broadband metro infrastructure that's been deployed here in the US -- and in other countries also -- that makes possible a new type of wireless network where you have less expensive base stations connected to this infrastructure. This broadband infrastructure can have a much higher bandwidth. This is going to enable a new set of handsets and much higher bandwidths. It will take time, but in this research organization it's our job to ensure that if and when these networks are deployed, they will work. Our interest is always to create innovations that really work. Not just general concepts, but actual products. That's what we're all about.