It was a day that would live on and on in computing history. Forty years ago, on Apr. 7, 1964, IBM (IBM ) introduced a groundbreaking computer, the IBM System/360 -- otherwise known as the mainframe, or simply big iron. Previously, computers tended to be single-purpose machines. But with the launch of the mainframe, governments, businesses, and scientists could use one machine for anything from plotting the trajectory of rocket launches to tracking insurance claims. And once they wrote a software program for one member of this family of computers, it would work on any other mainframe.
Like circuits on silicon, that day is etched on the memories of people who helped IBM CEO Thomas J. Watson Jr. introduce it. One of the fathers of the mainframe, Bob O. Evans, remembers standing in an IBM lab in Poughkeepsie, N.Y., explaining how the computer worked while TV cameras whirred. So caught up was Evans that "World War III could have started and I wouldn't have known it," he says. Indeed, it was a defining moment. Even today, a decade after pundits declared the mainframe dead, more than 70% of the world's digital information resides on the machines. And last year, IBM's sales of big iron actually increased 6%, to $4.2 billion, according to IDC.
What's more, the mainframe computing model -- the idea of tapping into powerful central computers -- has made a comeback. While PCs and small PC server computers remain important, many new tasks are now handled by powerful servers tucked away in data centers that serve the purpose mainframes did 40 years ago. Send an e-mail from your PC. Download music for your iPod. Order airline tickets on your Web-surfing cell phone. Somewhere, there's a muscle-bound server doing the heavy lifting.
Many of these machines differ from the early mainframes in one crucial respect. The original System/360 technology was proprietary, and once it became a de facto industry standard, IBM used it to beat competitors into submission and lock customers into its products. Microsoft Corp. (MSFT ) does the same today with its Windows operating system. But, increasingly, today's server computers operate on open-standard technologies, owned by no single company. It's a brand-new kind of machine. Businesses and consumers need not get locked into one company's products. And innovation should be able to flourish.
For a look inside the soul of the new machine, swing by a small lab at IBM's Thomas J. Watson Research Center in Yorktown Heights, N.Y. This is the birthplace of Blue Gene, a new kind of supercomputer that's essentially a cluster of servers. IBM is scheduled to turn the first version over to Lawrence Livermore National Laboratory in 2005. With 128,000 microprocessors strung together, it will be six times faster than today's fastest supercomputers.
Think of Blue Gene as the 21st-century mainframe. The company is considering installing smaller versions in supercomputing centers it recently opened in Poughkeepsie and France. The first uses are expected to include drug research and real-time pricing of financial derivatives. IBM foresees a time when doctors operating on a cancer patient will be able to take tissue samples, digitize the information, and send it to a remote computer for analysis while the patient is still in surgery.
In the boom days of the mainframe, this concept was called time-sharing. Companies would buy time on computers, paying only for as much as they used. That's now one of the foundations of a new vision called utility computing -- the idea that computing power should be as easy to tap into as the electrical grid. To realize this utility vision, though, today's servers have to become as reliable and secure as the first mainframes. IBM is transferring some of the core technology of its mainframes to other kinds of servers to make them more robust. At the same time, IBM wants its computers to fit in a world where open standards are starting to rule. So, IBM servers run the Linux open-source operating system as an alternative to its own software.
That switch to open standards may turn out to be the most significant change in the computing world for the next 40 years. The old-style mainframe is phasing out. Long live the new one.
By Steve Hamm