You've heard the rap. PCs can do this. Mainframes can't do that. PCs are cool and creative. Mainframes are plodding, ponderous and dull. PCs will make mainframes obsolete, dinosaurs of the Technologic Era. Mainframes will bring down IBM. PCs are the future.
Right?
Well, perhaps it's not quite so simple. Next to the Information Highway, a phenomenon known as downsizing has become the hot trend in computing. For those not up on their jargon, that's shorthand for a '90s vision of information management: switching from Big Iron to flexible networks of personal computers, and workstations. Technocrats call it client-server - myriad little computers busily swapping data and divvying up corporate chores that central mainframes used to perform. The benefits are supposedly huge. PCs are cheaper to buy, they say, and lots easier to use. They're great for linking workers to one another and the outside world. Just unplug that old-fashioned refrigerator, empower your employees with turbocharged PCs and watch productivity (and your bottom line) soar. The reality is far different, if you listen to the likes of Richard Finkelstein, president of Performance Computing in Chicago. Client-Server? Chaos. A nightmare. I've talked to hundreds of companies attempting the transition, and only one has fully succeeded.
By his reckoning, half of America's major companies have
been swept up by a fad that could waste billions. Too many
business people have been seduced by the hype of technology
revolution, adds Bill Laberis
(Footnote)
PCs save money. Say you're a CEO deciding whether to upgrade your old mainframe. A traditional mainframe might cost $2 million or more. For a quarter of that you could buy a network of PCs and desktop workstations. Suppliers assure you that the littler machines can do everything your mainframe does (and more). Add in the fact that they're anywhere from 10 to 100 times cheaper to run per unit of processing power, and the math looks pretty good.
Not according to a study by the Gartner Group in Stamford, Conn., which reports that buying and installing a PC network accounts for only 9 percent of a typical system's cost to the company. Factor in hidden expenses - writing coordinating software, retraining staff, adding support teams to maintain a decentralized information system - and PCs can end up costing twice as much as mainframes. International Technology Group in Los Altos, Calif., puts the figure even higher. A survey of 250 companies find that the cost of a mainframe per user averages $2,300 a year, compared with $6,400 for networked PCs. Switching to PCs simply shifts the costs from the information department to personnel, says Gartner's Peter Schay. Nobody should expect to save money.
PCs work as well as mainframes. Sorry, but they don't. Just ask yourself: how often do everyday users reboot their machines, or stumble over obscure software glitches? In a network, you can multiply these problems by hundreds if not thousands of desktops. Mainframes, by contrast, offer near-guaranteed reliability. Developed over the last 20 or 30 years, they are largely bug-free. By contrast, putting mission-critical jobs on PCs - a large corporation's inventory, say, or payroll account - can be disasterous.
Horror stories abound. Laberis tells of a leading investment bank, First Boston, that went network. Traders got their data faster (no small thing on Wall Street), but insiders say the new system cost nearly four times as much. Gartner's Schay tells of a major auto-leasing company that switched to PCs, only to find that it couldn't keep up with the crush of reservations. After spending millions of dollars, the company last year gave up (as happens in nearly a third of such efforts, according to the trade magazine Datamation) and went back to tried-and-true minicomputers. A senior executive at Unilever talks of his nerve-racking five years; work to make the transition, as yet without success: People talk as if this technology were here. Just plug in and play. But it's not 'here.' It won't be for a very long time.
Such disillusionment has the folks at IBM practically crowing ing. The mainframe won't go away, says William Reedy, a marketing director for Big Blue's heavy metal. Revenues from big computers are trending down, but that's largely because of aggressive (near 50 percent) price-cutting. Volume is up sharply. Says Reedy: We can't keep up with demand.
None of this means that networks of PCs are a dead end. To the contrary everyone agrees that's the way the industry will gradually go. But a new sense of realism is clearly taking hold. For the next decade, at least, the computing world will be a hybrid. Because PCs offer clear advantages - speed, flexibility, and enormous utility as a creative tool - they will take on more and more jobs in the workplace. But for big tasks in big corporation - where large amounts of data need to be stored, transferred or processed - heavy metal will rule. Mindful of this Richard Shaffer at Technologic Partners in New York offers some advice to anyone thinking of unplugging his mainframe: think twice. The humming-thrumming Big Thing in the back room may be a dinosaur, but it's nowhere near extinct.
Michael Meyer