IBM's Gamble In The 60's Continues To Reap Rewards

Forty years ago next week, IBM unveiled the results of a multibillion-dollar bet that transformed the world of computing. The company overcame false starts and in-fighting to usher in a new era of technology by introducing the System/360, the first IBM mainframe.

IBM gambled its future, dedicating an unprecedented $5 billion for the development and manufacture of a new line of computers that would eventually commercialize such staples of the modern computing world as system compatibility and transaction processing. The investment, which would equal tens of billions of dollars today if adjusted for inflation, exceeded what the U.S. government spent on the Manhattan Project to develop the atomic bomb in World War II.

Fred Brooks Jr. was given the responsibility of creating a new high-end platform for the Data Systems Division, and his group designed the 8000 series. After a presentation to IBM management in January 1961, the 8000 series was rejected by Learson and Evans, who believed that any new product line ought to unite the Data Systems and General Products divisions under a common platform. The decision to kill the 8000 series devastated Brooks and his division. "We fought for six months whether to go with the 8000 series or some vague, ill-defined, new product line," Brooks says.

Brooks saw that IBM was losing ground at the high end of the market to competitors, in part because of the Stretch failure. He believed a new 8000 series could be in place by the end of 1961 to revitalize IBM's efforts. A completely new product line, particularly one as ambitious as the one proposed by Learson and Evans, would likely not be ready until 1964 or 1965. "We thought we needed to do something right then," he says.

The divisions were at war. Brooks and the Data Systems Division decided to move forward with the 8000 series on their own. "When I went to Poughkeepsie in early 1961, they were lined up four-square against me," Evans says. "It wasn't exactly hate, but it was fierce competition."Evans, with Learson's support, won the battle, and the 8000 series was scrapped. IBM began development of the NPL, which stood for new product line. In a surprise move, Evans named Brooks to head up that line.

In November 1961, Learson set up the Spread task force to determine the parameters of the new product line. Compatibility was a key criterion. The task force moved to create a line of computers that would be both upward and downward compatible, meaning that programs developed for the low-end machines would run on the big computers and vice versa. "It was one thing to build a small, affordable machine," Evans says, "but quite another thing to build a supercomputer that had the same instruction set that allows each and all [computers] to run the other's programs."

Challenges included shifting from 6-bit bytes to 8-bit bytes, which would open up text processing as a computing application but also require a new scheme for labeling and routing signals as they traveled through the computer, and determining what type of circuitry technology would be used, Evans says.

To take full advantage of system compatibility, IBM saw that it also needed to provide a wide variety of compatible peripheral devices and equipment. That required creating a standard input/output interface, a concept first considered for Stretch. With the introduction of 360 in 1964, IBM simultaneously introduced 144 peripheral devices such as printers and keyboards that could be used throughout the product line.

Key to hitting the price-performance goals necessary to ensure the success of the 360 was moving to an advanced circuitry technology. The state of the art then was called Standard Modular Systems, which used individual resistors and capacitors attached to a circuit board. On the horizon was the integrated circuit, which eventually would revolutionize electronics."It was clear we needed something that could provide a reduction in cost in an order of a magnitude, and an improvement in performance in an order of magnitude," Bloch says. "And we needed an increase in circuitry density by a factor of 10 to 100, and that was just not available."

The decisions made by IBM executives beginning in 1961, and culminating with the formal introduction of the 360 series on April 7, 1964, would cement the company's position at the forefront of the computer revolution of the last half of the 20th century.

"It was clear we were taking a major risk," says Erich Bloch, who led development of microcircuit technology for IBM in the early 1960s. "It was clear that this was all or nothing."

"If the 360 didn't succeed, we knew we were dead in the water," says Bob Evans, who led the project as development VP for IBM's Data Systems Division. "The competition would have rolled over us. But we also knew the market was wide open for innovation, and while IBM at the time was emerging as a leader, we certainly didn't have command of the market."

In 1961, computing remained more science fiction than everyday reality. Fewer than 30,000 computers were installed worldwide, and most of them were used by the scientific community and government agencies. Circuit technology had only recently moved beyond vacuum tubes to discrete transistors placed on boards at a density of a few thousand per cubic foot. "Supercomputers" with less capability than a modern laptop sold for many millions of dollars, and most systems were rented, not sold.Early attempts at a high-powered machine with broad appeal were marked by internal rivalries and commercial failures. Thomas Watson Jr., who succeeded his father as chairman of IBM in 1956, gave T. Vincent Learson responsibility for IBM's computer-development and -manufacturing operations, which were divided into two divisions. The General Products Division made lower-end computing products that rented for about $10,000 a month or less, and the Data Systems Division produced higher-end systems that rented for more than $10,000 a month. The General Products Division, in Endicott, N.Y., had recently scored big with the 1401, the first computer with more than 10,000 installations. The Data Systems Division, in Poughkeepsie, N.Y., made the more-powerful, but low-volume, 7000 series. A natural rivalry between the two groups came to a head over the development of the 360 series.

The Data Systems Division designed the ambitious "Stretch" computer as its platform for the 1960s, spending about $30 million in development. At a price tag of more than $13 million each, Stretch attracted few buyers and was killed off with little to show for IBM's investment. Pressure was intense to ensure that the next effort targeting the high-end market was a success.

Although integrated circuits were seeing limited use in defense programs, IBM determined that within the time frame it was targeting for the 360 series, integrated circuits wouldn't be available in the volume and price range it needed. IBM decided to move forward with Solid Logic Technology, which placed individual transistors, capacitors, and multiple diodes on a chip that was encapsulated in glass, attached to silk-screened wiring, and housed in a ceramic package placed on a board. Going with Solid Logic Technology required the development of new manufacturing processes, as well as testing and assembly procedures, Bloch says. IBM also had to build new manufacturing floors.

As development went forward on the 360, those involved met on a regular basis to review progress and address requests for changes in the development manual created for the project. With a deadline of early 1964, representatives of the various groups would meet regularly in 1962 and 1963 to resolve as many as 100 outstanding issues at a time, Brooks says. "My experience with IBM is there is always panic going on," he says. "With each particular problem that came along, there was a high degree of action required."

The worries continued after the System/360 was launched in April 1964. Sales started off slow and didn't pick up until the machines had been on the market for a few months.In 1962, IBM's revenue was $2.6 billion. In 1964, it jumped to $3.2 billion, growing to $5.3 billion by 1967 and more than $7 billion by the end of the decade. System/360 sales far exceeded the company's projections, and IBM was unable to meet demand, in part because its manufacturing operations couldn't produce enough high-quality components. The compatible operating system also proved to be difficult, and a number of the glitches weren't resolved until the fourth release of OS/360 in November 1965.

Customers, however, began to see the advantages that could be obtained with the 360, and rivals felt the impact. IBM's competitors at the time--Burroughs, Univac, NCR, Control Data, and Honeywell--were known in the industry as "the Bunch." They never recovered from the introduction of the 360, says Mike Kahn, an analyst with the Clipper Group, a technology acquisition consulting firm. "The 360 has had a bottom-line impact on the evolution of computing and the creation of information technology as we've looked at it the past 30 years," Kahn says.

Irwin Sitkin, who developed the first information systems for Aetna Insurance in the late 1950s and early 1960s, says the 360 was instrumental in transforming the insurance industry. "If there were no [360 series], there would be no business," Sitkin says. "It was not trivial and not without risk, and the miracle of it all was it worked. It changed the nature of the way we operated from being merely accounting and statistics to true online policy creation and service."

Aetna and American Airlines were among the first to deploy the 360. American Airlines used the computer to expand its Sabre reservation system. Aetna created Safari, the System for Automatic Family Automobile Renewal and Inquiry. Aetna recruited personnel from throughout the company to build a technology workforce of about 200, who then proceeded to write more than 2 million lines of code to run on the 360.

In 1965, Aetna used one high-end 360 machine to handle automotive polices written in Vermont, "where cows still outnumbered cars," Sitkin says. The transaction-processing capability of the 360 let Aetna start automating the process of creating an insurance policy. Until then, field representatives had mailed handwritten policies to the home office, where nearly a quarter of them had detectable errors. Using the 360 running with a then-impressive 15-second response time, field representatives used remote terminals to file relevant policy information directly to the home office, where coding and calculation of premiums were completed and responses dispatched to branch-office printers. The results were so positive Aetna quickly purchased six more machines at a cost of around $3 million each and expanded the automation process to include all its automotive, homeowners' insurance, and workers' compensation policies. Aetna still uses mainframes today. "I don't think it's an overstatement to say the mainframe has been the engine that has driven the business-computing evolution," says Charles King, an analyst at the Sageza Group, a technology research and consulting firm. "They set the table for what has resonated for a long time."Bill Zeitler, current senior VP and group executive of the Operating Systems and Technology group at IBM, says the 360 "was a once-in-a-generation kind of statement. But if the mainframe only did today what the original system did, or its successors did, it would be dead already."

The mainframe has been able to stay a vital part of IBM and the IT industry because it has evolved and expanded its capability to include new technology platforms such as Linux, Java, and WebSphere, Zeitler says. While the 360 series defined the mainframe class of computers with its ability to support multiple users and the compatibility and scalability to meet customer requirements, today's mainframe is defined as a high-end computing environment that uses virtualization to support multiple operating environments, he says.

Rivals have been trying kill off the IBM mainframe for decades. For two years, Sun Microsystems has been offering a "rehosting" program that moves companies off mainframes to its Sun Fire servers; to date, Sun has worked with about 300 companies, says Don Whitehead, director of mainframe migration for Sun. "The mainframe has been a huge success story in the marketplace," he says. "A 40-year run for any technology is just tremendous. But we think maybe the mainframe should start thinking about a retirement strategy."

IBM believes the mainframe isn't ready for the rocking chair. The most recent addition to the five decades of IBM mainframes, introduced last year, is the Z990, code-named TRex. "Some of our competitors were saying the mainframe was a dinosaur, so we gave the new systems code-names of carnivorous dinosaurs," Zeitler says. "If you think we're a dinosaur, you're going to be rudely surprised."