A UCLA Professor and Net Pioneer Paves the Way for the Next Big Thing
Leonard Kleinrock is getting ready for the era of ubiquitous computing
By FLORENCE OLSEN
Santa Monica, Calif.
"Oh, you want to get to D? Well, next you go to C -- hop, hop, hop." The man making the sound effects is Leonard Kleinrock, whose pen is busy rapidly drawing loops and lines on a scrap of paper. He is explaining how packets of data are routed inside a computer network.
Well into the second of his two careers devoted to Internet research, Mr. Kleinrock, who is 65, can't stop teaching -- or inventing. Although he is "semiretired" after 36 years as a computer-science professor at the University of California at Los Angeles, he is still advising graduate students and working on new algorithms for network communications.
In the 1960's, Mr. Kleinrock was the first to describe the queuing principles still used today in designing every type of computer network, from a campus Ethernet to the global Internet. He is also one of the few engineers from the Internet's early days who opted for a university career.
At U.C.L.A., where he is credited with producing dozens of sought-after Ph.D. graduates, Mr. Kleinrock's dedication to teaching is legendary.
"I never planned to teach, but I just loved it," he says.
"I'll tell you all the things I love about being a professor," he says without a pause. "You're working with young people -- keeps you fresh, you have to stay sharp. You have no boss. You do research in any field you like. You travel a lot. The best minds in the world come to the university to lecture or visit. It's a prestigious job. The salary is not bad. If you work harder, you can get more. But it's the independence that I really liked."
Mr. Kleinrock is busier in semiretirement than most people are when working full time. On days when he is not at U.C.L.A., he is behind a desk in his sunny office here, having given up the title "department chairman" to become a corporate chairman instead.
Joel Short, one of his recent Ph.D. graduates, is down the hall. Teacher and former student have formed a company, Nomadix, to invent and manufacture communications devices for the next wave of Internet use.
Mr. Kleinrock and Mr. Short are paving the way for what they say will be the Internet's next big thing, what Mr. Kleinrock and many others call nomadic computing -- and the next big thing after that, which is ubiquitous computing.
"Let me give you the vision thing first," Mr. Kleinrock says. The vision thing is his "farthest-out idea of where the Internet is going."
The Internet will be available everywhere -- and always "on," he says. Sensors will be embedded throughout the physical world and connected wirelessly to the Internet, or whatever its successor is called in the future.
"Walls and desks will be full of actuated sensors, logic, memory, processing, cameras, speakers, microphones, displays, and communications," he says. "So when you walk into a room, the room will know you just walked into it, and you can talk to the room and say, 'I'd like some information about The Chronicle of Higher Education.'"
Between where computing is now and that ubiquity, he says, will be nomadic computing: Wherever people go, they will be able to gain access to information over the Internet using a variety of handheld devices.
Mr. Kleinrock's company is designing new algorithms, or instructions, which will go into network hardware that lets people stay connected to the Internet, no matter where they are.
For anyone interested in the Internet, these are exciting times, Mr. Kleinrock says. If fewer computer-science and business undergraduates are choosing to go on to graduate school, who can blame them? "It's a gold rush," he says. Some students simply feel "the opportunity cost is too high to spend another year or two in graduate school -- in Internet time, that's an eon."
The lure is more than just money, he says. "They're looking for positions in start-ups. They're willing to take a chance."
In his own career, he has never shied away from taking chances. Mr. Kleinrock graduated from the Bronx High School of Science in 1951, and earned his bachelor's degree in electrical engineering by taking evening classes at the City College of New York. After graduating at the top of his class, he earned a scholarship to attend the Massachusetts Institute of Technology, where he received a master's and a Ph.D. in electrical engineering.
Today, he enjoys an unofficial role as historian for the small group of engineering scientists who, 30 years ago, were his colleagues in building a network that was the Internet's precursor.
Mr. Kleinrock was in graduate school at M.I.T. when he got to know Claude E. Shannon, the mathematician who discovered how coding could be used to send messages reliably over noisy communications lines.
"This was a great mathematician" -- and a person who had a knack for seeing the connections between physics and mathematics, Mr. Kleinrock says. He recalls that he once found Mr. Shannon, who was teaching at M.I.T., using a Swiss Army knife to take apart a differential -- a complex arrangement of gears.
"He's still my role model," Mr. Kleinrock says.
In the late 1950's, what remained to be solved in Mr. Shannon's field of information theory "were hard problems of relatively small consequence," Mr. Kleinrock says. Small challenges didn't appeal to him, he says, "so I started studying data networks.''
"I've always enjoyed thinking about large systems with thousands or millions of participants," whether it is the communication that can be observed in an ant colony, or the communication between computers talking to each other, he says.
Soon, Mr. Kleinrock was using a type of probability theory known as queuing to understand the optimal ways of sending and receiving messages inside large computer networks -- which, at that point, didn't exist. The largest network that existed at the time was the telephone system, but it lacked the engineering characteristics that drew his interest.
The basic data-network principles that Mr. Kleinrock proposed are now familiar to network engineers, but at that time they challenged what everyone understood about large networks. Queuing-system equations were also extremely difficult to solve mathematically. "I came up with a model you could solve by making the right assumptions," Mr. Kleinrock says.
Mr. Kleinrock described these principles in 1961 in his dissertation, published by McGraw-Hill as Communication Nets and widely regarded as a seminal work in data-communications theory. He also discussed his discoveries with Lawrence G. Roberts, his office mate at M.I.T., who was soon in a position to give Mr. Kleinrock's assumptions a reality test.
In December 1966, Mr. Roberts began working for the Defense Department's Advanced Research Projects Agency, where he became the chief scientist for the Information Processing Techniques Office. One of his first projects at ARPA was to design and build a computer network to allow research scientists under contract with ARPA to gain access to each other's specialized computers.
ARPA's research computers at U.C.L.A. were used for simulation and modeling, while those at the University of Utah were customized for graphics processing, and computers at other ARPA sites had other specialized capabilities. As projects grew more complex, the researchers at various institutions started asking for computers like those the government had already bought for their colleagues elsewhere.
Mr. Kleinrock says the government's reaction was: "Whoops, we can't afford this, so let's put you on a network."
"Nobody cared about data networks," until the government found a need for them in the mid-1960's, Mr. Kleinrock says.
Because of Mr. Kleinrock's theories, Mr. Roberts says, he was confident that computers could be connected. He would design the network using the principles of decentralized control and "access-on-demand" queuing, which Mr. Kleinrock had first described to him in 1961.
But in 1967, people in the communications industry told Mr. Roberts that he was crazy to propose investing millions of taxpayer dollars to build a network using unproved packet-switching technology. "Because Len had done that work," he says of Mr. Kleinrock, "I could stand up to those people."
Mr. Roberts asked Mr. Kleinrock to help write the specifications for the network proposal that ARPA put out for bid, and in 1969, Bolt, Beranek, and Newman Inc., of Cambridge, Mass., began building what became known as the Arpanet. By then, Mr. Kleinrock was at U.C.L.A. and had signed a contract with ARPA to record and analyze the internal workings of the fledging network.
At U.C.L.A., the government installed a refrigerator-sized version of a device that would today be called a router, making the U.C.L.A. campus the site of the Arpanet's first node. It's still there. "It's just down the hall in the computer-science library, in a corner of the room, not at all showcased, unfortunately," Mr. Kleinrock says. After the device was installed, in 1969, the Arpanet grew rapidly, eventually evolving into the global web of networks that is the Internet today. Mr. Kleinrock subsequently wrote a pair of books on queuing theory and its applications for engineers that are now classics in the field -- Queueing Systems: Theory and Queueing Systems: Computer Applications (John Wiley & Sons).
Federal research money flowed freely to the pioneers of the Internet and resulted in "a huge amount of progress," Mr. Roberts says. Because of free and open access to the basic documents and protocol specifications for the Internet, he says, the network expanded rapidly. With the money he received from ARPA, Mr. Kleinrock supported a large group of U.C.L.A. graduate students who made significant contributions of their own to the creation of the Internet.
One of Mr. Kleinrock's students was Vinton G. Cerf, a member of the group that wrote the network-control software for the Arpanet, and who later, with Robert E. Kahn, wrote the communications software that became the standard TCP/IP protocol used on the Internet. "It was an extraordinary time for most of us," recalls Mr. Cerf, who is now the senior vice president of Internet architecture and engineering for MCI WorldCom Inc. Mr. Kahn is now president, chief executive officer, and chairman of the board of the Corporation for National Research Initiatives.
Throughout his years at U.C.L.A., Mr. Kleinrock says, he received ample money from ARPA for network research, approaching $1-million a year. Since 1990, however, Mr. Kleinrock has pared his ARPA research "down to zero," and he is now working with only four graduate students.
Mr. Kleinrock has also influenced public policy related to the Internet. As the chairman of a committee for the National Research Council's Computer Science and Telecommunications Board, he oversaw the writing of a 1988 report, "Toward a National Research Network." The report was widely read in Washington and influenced the passage of the High-Performance Computing Act of 1991, which established the National Research and Education Network.
Mr. Kleinrock has received dozens of prizes and honors, among them the prestigious L. M. Ericsson Prize in telecommunications, the Institute of Electrical and Electronics Engineers' Harry M. Goode Memorial Award, and numerous teaching and faculty awards.
"Lennie's spirit and energy level" are remarkable, says Howard Frank, the dean of the Robert H. Smith School of Business at the University of Maryland at College Park, who has known Mr. Kleinrock for 35 years.
Mr. Kleinrock likes to think just for the fun of it, says Mr. Frank. He remembers traveling to a professional meeting with Mr. Kleinrock and Mr. Roberts. The two of them "drove me crazy," he says, challenging each other with ridiculously complex mathematical word problems and solving them in their heads while they took turns driving.
Mr. Frank also remembers an incident from the mid-60's, when he was teaching at the University of California at Berkeley and Mr. Kleinrock was teaching at U.C.L.A. From time to time, the two friends gave guest lectures in each other's classes.
On one of those occasions, Mr. Kleinrock had been camping in the mountains when he blew a tire coming down a mountain road. After a desperate, zigzagging descent, he got his car down the mountain, hopped into a small plane, and made it to Berkeley in time to give his lecture on computer networks. He illustrated his talk using glass slides, every one of which was cracked.
"He was just as great as if he had never gone through that at all," Mr. Frank recalls. "He's a fabulous teacher." Indeed, over the years Mr. Kleinrock has been the adviser to 43 computer-science Ph.D. graduates, some of whom are now teaching at Columbia, Stanford, and Tel-Aviv Universities, and elsewhere.
"The number of students he has turned out is enormous," says Mr. Roberts, who recently hired one of those Ph.D.'s at his own Internet start-up company, Packetcom, in Palo Alto, Calif. Mr. Roberts needed an expert in queuing theory, he says, and, naturally, he thought of Mr. Kleinrock.
"Very few people in the world teach queuing theory," he says. "If you're in that area, you go to Len."
http://chronicle.com
Section: Information Technology
Page: A53
