Profile: Irving Wladawsky-Berger

Well, Somebody's Got to Reinvent the I.B.M. Mainframe

By Steve Lohr
The New York Times

September 12, 1993

Irving Wladawsky-Berger tugs a silver handle on a black drawer that is somewhat bigger than a shoe box and is made of latticed steel. Inside, colored wires and computer chips are attached to a couple of circuit boards. Two little fans stand ready. Each unassuming drawer, Mr. Wladawsky-Berger explains, holds the innards of a computer work station, and the caged work stations are housed in tall racks.

It may not look like much, but it could be I.B.M.'s future, or a large part of it anyway.

Perhaps the biggest challenge facing the International Business Machines Corporation is to make the transition from its traditional mainframe business to a form of industrial-strength computing based on low-cost microprocessor technology. The task amounts to reinventing the mainframe, a daunting effort being directed by the 48-year-old Mr. Wladawsky-Berger (pronounced Law-daw-skee Berger), general manager of the company's supercomputer unit, I.B.M. Power Parallel Systems.

His is not just a big job, it is a delicate job. For all the talk of the death of the mainframe computer, it remains a huge business for I.B.M., which had more than $20 billion of revenues last year. Though far less lucrative today than it was a few years ago, the mainframe business is still a cash cow, and I.B.M. must milk it for as long as possible.

The mainframe business supports many thousands of I.B.M. jobs and, one executive noted, there are still hordes of "mainframe bigots" in the company. To them, Mr. Wladawsky-Berger looks like a cannibal -- even though he is a small, bespectacled man with curly graying hair, holds a Ph.D. in physics from the University of Chicago and maintains a professorial air about his person.

Still, the technology, the marketplace and even I.B.M. now seem inevitably headed Mr. Wladawsky-Berger's way.

"It's a lot easier to influence the business when things are not going so well," he said recently during a tour of the supercomputing lab at I.B.M.'s research center in Yorktown Heights, N.Y.

Despite hurdles, especially in software development, the shift to big computers that run on microprocessors is a given in the industry. Only the timing of the transition is in doubt.

Parallel machines link dozens or even thousands of microprocessors, the computers-on-a-chip that are the brains of work stations and personal computers. In contrast, traditional mainframes run on a single big processor that uses so much power and throws off so much heat it is water-cooled. Microprocessors are mass-produced, while the old bipolar processors are essentially handcrafted.

So the microprocessor-based machines deliver computing power at about one-tenth the cost of mainframes, analysts say. "This is where the world is headed, and Irving is a realist, tuned into the change, not looking at the world from a one-way I.B.M. window," said Gary Smaby of the Smaby Group Inc., a consulting firm in Minneapolis.

Power Parallel Systems is less than a year old, and its few sales have thus far been mainly multimillion-dollar machines sold to advanced research labs like the Cornell Theory Center, a supercomputing site. Earlier this year, the Cornell center bought an I.B.M. SP-1 machine with 64 processors linked together, in parallel. Last month, Cornell announced that it would scale up the I.B.M. machine to 512 processors (512 of the big, shoe box-size drawers), helped by money from New York State, the Department of Defense's Advanced Research Projects Agency and the National Science Foundation. But companies like Halliburton, American Express and American Airlines are looking closely at I.B.M.'s parallel processing machines, and some corporations are already experimenting with them.

Mr. Wladawsky-Berger spends much of his time these days meeting with companies that use high-powered computers. His goal is to sell nearly 100 parallel machines by the end of this year, and he wants to ship "hundreds" in 1994. Most of them, he predicts, will be 8-processor and 16-processor machines, costing from $300,000 to $1 million.

"We're not building specialized machines for just a few research labs," he stressed. "We're making machines for the commercial market."

For at least a couple of years, however, the commercial market will be confined to companies that need leading-edge computing, like Halliburton Energy Services Group in Houston. The oil-services company has been experimenting with an I.B.M. parallel computer for six months to help it get faster, more refined, three-dimensional pictures of oil-bearing geologic formations. In the high-tech hunt for hydrocarbons, powerful parallel machines are a tool to save time, cut costs and drill fewer dry holes.

"The days of the mainframe computer in our industry are numbered," said Walt Ritchie, Halliburton's manager for geophysical computer processing.

In Mr. Wladawsky-Berger's view, the promise of parallel machines is to bring supercomputer firepower, inexpensively, out of the labs and into the mainstream of corporate America, allowing people to store and analyze vast amounts of data, whether it is collected "from satellites in the sky or supermarket scanners," he said.

A few examples: Retailers tracking point-of-sale data will be able tailor their offerings on a store-by-store basis. Pharmaceutical companies will be able to shorten drug-development times by doing ultrafast calculations to simulate molecular interactions. Hospitals will be able to store, sort and analyze patient data, enabling faster and more accurate diagnoses. Airlines, doing fast computations of weather data, will be able to make last-minute adjustments in flight paths to save fuel.

It is the microprocessor revolution -- which exhibits the rapid pace of price and performance gains and the same relentless advance that threatens the mainframe world -- that is putting the myriad new applications within reach.

"Personally, I'm convinced that replacing the mainframes is the most boring side of what these machines will do," Mr. Wladawsky-Berger added. "It's the new applications that are really exciting and, ultimately, the much larger opportunity."

Besides, it will take some time to replace the I.B.M. mainframes -- and software is the obstacle. Businesses and governments worldwide have invested an estimated $1 trillion in mainframe software, for everything from corporate bookkeeping and payroll ledgers to Federal tax records.

YET the real issue for I.B.M. and its rivals is how quickly the mainframe will be supplanted, and which companies will lead the transition. Not surprisingly, the debate on that subject is lively, in the industry as well as inside Big Blue itself.

According to the International Data Corporation, the market for parallel machines, for scientific as well as corporate uses, is modest but growing rapidly. That market reached $583 million this year and, I.D.C. estimates, it could reach $1.6 billion to $2.6 billion by 1997, depending on how rapidly major corporations embrace the technology. At the Gartner Group, Howard A. Richmond, director of high-performance computing research, said the market for parallel machines could reach $5 billion by 1997. "This market is poised to absolutely explode," he said.

For its part, I.B.M. seems to be hedging its bet on parallel computing. Later this year, the company plans to introduce a so-called 390 Parallel machine, essentially a mainframe with a parallel processing turbocharger. And the Power Parallel unit is working on vastly increasing the storage and data-retrieval capabilities of I.B.M.'s microprocessor-based machines -- the two traditional strengths of mainframes.

Ultimately, the goal is to have a set of almost interchangeable, mix-and-match big machines from I.B.M. that are able to run its old software and new applications.

How touchy is the timing issue for I.B.M.? In an interview, a colleague of Mr. Wladawsky-Berger said I.B.M. should have machines to "cover the entire market in a couple of years."

The next day, a spokesman for the mainframe business called to say that I.B.M.'s official line is that it might take longer: the company plans to have a microprocessor-driven "general purpose" machine -- a mainframe replacement -- by the "second half of the decade."

That is less a contradiction than a clarification, but any suggestion that I.B.M. can actually manage to keep up with the market's shift to the post-mainframe world amuses Michael Lambert, the general manager for the NCR Corporation's large-computer-systems division in San Diego.

NCR is the clear leader in the nascent commercial market for parallel computers, with estimated sales this year of $300 million. Any delays or mixed messages from I.B.M., Mr. Lambert said, are marketing plusses for NCR. "The business we're getting is coming at the expense of I.B.M.'s mainframe business," he said. "It's an enormous opportunity for us."

It represents a big challenge for Mr. Wladawsky-Berger, however. But he has succeeded in making some difficult transitions in the past, not just professionally, but personally.

He was born and raised in Cuba, the son of Jewish Eastern European immigrants. His hyphenated last name was "part of trying to be Cuban, strange as that might seem now," he said, referring to the practice in Spanish-speaking countries of children taking the surnames of both their parents. (Berger is the family name of his Polish mother, Rachel, and Wladawsky comes from his Russian father, Julius.)

AFTER Fidel Castro came to power in 1959, Mr. Wladawsky-Berger's father, a store owner, lost everything, and in 1960, at 15, Irving and his family fled to Chicago, where they had relatives. He quickly learned English (he has kept a lilting Spanish accent) and became an outstanding student.

Today he lives in Westport, Conn., with his wife, Cheryl, and two children, Zara and Ethan. And Mr. Wladawsky-Berger credits Cuba with giving him a lifelong affection for baseball. He speaks sadly of the New York Mets' having a "tragic" season this year, and he has never forgiven George Steinbrenner, owner of the New York Yankees, for trading Reggie Jackson. He is also an avid reader of Elmore Leonard's crime thrillers. "I love the scenes of Miami lowlife," he said.

After earning his doctorate at the University of Chicago, he joined I.B.M. in 1970 as a researcher at the Thomas J. Watson Research Center, a white-coat setting at the time, and an area well removed from the world of products.

During a one-year sabbatical in the marketing division in 1976, however Mr. Wladawsky-Berger found he liked the business side of things. In 1985 he became vice president for development in I.B.M.'s mainframe division, where he worked on technical strategy.

Early on, Mr. Wladawsky-Berger began advocating the use of parallel machines. Still, he admits he did not foresee how fast microprocessor technology would advance and eat into the mainframe world.

Within I.B.M., his strength is regarded as being cross-cultural: he is able to talk business to the company's executives, and he can talk science to its researchers.

"He's a visionary thinker who can also sell that vision both to customers and to other I.B.M. executives," said Rick Baum, an I.B.M. Fellow who is also the director of systems strategy for mainframe business.

Asked for the sound-bite version of his pitch, Mr. Wladawsky-Berger replied: "This is the future -- and I.B.M. will ride it. And we will work with you to get to the new world."