March and April are the best months for kite-flying. The balmy sunshine still seems to be a special gift. The breezes are strong, if capricious, and they smell wonderfully fresh. After a long and difficult winter, we look forward to the crowd of kite enthusiasts that, even in New York, fill the parks. So it seemed particularly jarring to discover that in Mountain View, California, the very heart of the high tech district known as Silicon Valley, kite-flying is one of the most hazardous of activities. According to the news wires, DeAndra Anrig, an eight-year-old known as Andi, went out to fly her kite on the 24th of March. Things being dicey in semiconductor country these days, somebody apparently thought 73-pound Andi ought to be securely tethered to her flying object. Some 200-pound-test fishing line just happened to be available. As little Andi launched her kite high into the heavens, a small aircraft was rushing a patient to nearby Stanford University Hospital. The low-flying plane flew smack into the kite string, which turned out to be the most underrated piece of plastic in the Western Hemisphere, if not the galaxy. The string caught on the plane’s propeller and, according to the press, Andi became airborne. Furious Before she could figure out what to do, little Andi had travelled an estimated 200 feet at an altitude of approximately three metres. The onrushing silhouette of a tree encouraged the child to loosen her grip on the string, and the rest was a matter of gravity. Interviewed by an earnest United Press International reporter, Andi’s mother Debbie reportedly said of her aerobatic daughter, She’s just very, very sore. Had it been us, we would have been furious. Users of IBM mainframes will have no trouble identifying with DeAndra Anrig. For years, they have tried to fly with the latest mainframes and disks their companies could afford. By one or another means, they have stretched budgets, fiddled with financings, conned IBM out of OSTAs and VPAs, scrimped and saved. And the very minute they’re ready to rush into the boardroom and show off their handiwork, IBM springs new software, hardware, pricing or policy that carries these poor souls, like little Andi, in the general direction of the woods they have so recently escaped. For most users, the toughest strategic challenge this year will be posed by IBM’s whirlwind changes in the 3090 line up. The company has promised a new MVS operating system, at the same time threatening users with yet another alternative. Fortunately for most large shops, AIX, as IBM calls the Unix of the Armonk Palace, is a year away. Besides, AIX needs to run under VM, a distinguished IBM software product that most data processing users have learned not to appreciate. Fully aware that it will sell MVS/ESA to every mainframe shop, IBM has vowed to make every one of its installed 3090E systems conform to the hardware revisions that the software will require. This tinkering will have an interesting side effect. It will make it easier for IBM to announce other changes in the 3090 series. We believe that the two most significant possibilities this year are a performance kicker and a fast memory option. The memory upgrade is in some ways a more interesting possibility than the F model performance kicker. There is no question that computers can do more work when they have larger or faster memory. Not only does better memory enhance a system’s processing rate, but it also can accelerate the rate at which data is transferred.
In other words, faster memory means a complex can include more processors, faster processors, more channels, and faster channels. For the past few years, IBM has been deservedly proud of its progress in memory chips. Yet the company has delivered only a small portion of the benefits its chip developments have promised. IBM mainframe memory is terribly expensive but not impressively fast. IBM’s engineers are undoubtedly aware of this. We strongly suspect that they have looked into the possibility of providing significantly faster memory for some, if not all, 3090s. The res
ults could well surface in the form of products announced concurrently with faster 3090 processors. A new generation of main memory would have an immediate and pronounced effect on every big IBM user, on every lessor, and on IBM itself. Suppose IBM announces 3090F machines that run, say, 10% faster than E models using current main memory, but 20% to 25% faster when they are fitted with new memory. One can assume that the performance benefits of the new memory would be greater on larger machines – the 300F, 400F, 500F and 600F. The largest users would have to move to such new memory. Whether or not 3090s could be configured with a mix of normal and accelerated memory, most shops would trade up. Users that own multiple 3090s might shift some memory between their systems. But most users would try to unload the old memory, much the way they de-install unwanted disk drives or printers. This could play to another IBM advantage. Most independent refurbishers are not equipped to handle a large volume of on-site memory upgrades; they do most of their mainframe reconfigurations in their factories. IBM, on the other hand, would be delighted to have the work. Unlike the more technical processor upgrades and service jobs that sometimes strain even IBM’s considerable resources, memory upgrades are done by the book. Now that all the 3090E machines are being equalised to provide ESA services, IBM could conceivably train and reassign technicians to its large mainframe squad from other divisions. This crew would be dedicated to providing timely memory upgrades. Unlike third parties, IBM can withstand some criticism if a greenhorn technician fails to make a memory upgrade work the first time. We foresee this eventuality because we believe the IBM mainframe market is ripe for change. Besides, there is an abundance of evidence that IBM has generally adopted a strategy of announcing frequent model changes. Like the automobile companies, IBM has apparently determined that users like the idea of progress. Users will, for the most part, buy improved machines as fast as IBM can provide them. Turmoil They seem to be coping with the turmoil such frequent change brings. For its part, IBM has found a way to moderate the swings in revenue that its sales-only policy would produce every generation. In effect, IBM has switched from a generational strategy to a feature strategy. IBM’s big guns may not really want to run it this way. But they are beholden to the shareholders, who are interested only in the rewards, not the risks, of high technology. Unfortunately, IBM’s shareholders, managers, and customers can’t get one without the other. We think IBM is hoping for an excessive reward in the mainframe business this year, and hopes to achieve it by getting users to take the risks. ESA is still vapourware, and new mainframes are very expensive. Unless business conditions turn out a lot better than they now look, users could tell IBM to go fly a kite. They won’t wish on the company an experience like poor little Andi had, of course. But little Andi didn’t plan to get carried away, either. Reprinted from Computer & Communications Buyer: details on request.