One of the myths of information processing is that mainframe shops are not very good at shopping for systems. This myth is widely believed by personal computer and Unix users who wouldn’t know a large-scale computing job if it chomped on their coccyx. It is not a belief held by many people who have tried […]
One of the myths of information processing is that mainframe shops are not very good at shopping for systems. This myth is widely believed by personal computer and Unix users who wouldn’t know a large-scale computing job if it chomped on their coccyx. It is not a belief held by many people who have tried to sell or lease machinery to the big players, not lately, anyway. There are plenty examples of data processing managers signing for equipment that seems luxurious by the standards of small business. But far more often than not, costly systems at large enterprises provide benefits that are otherwise unavailable, such as reliability, resiliency and residual value. To gauge the buying patterns of big companies accurately, you have to look at the choices they make when they get the exact same equipment used by smaller and purportedly savvier companies. Personal computer servers provide such an example. At mainframe sites, which include not only users with the latest IBM Corp 9672s but also shops still waiting for their last 3090 lease to end, server buying patterns favor fast, affordable machines from trustworthy vendors.
IBM is only one supplier and when it comes to Pentium boxes, it is not even the leader. Compaq Computer Corp, with its reputation for quality and performance, is the number one vendor. IBM is number two, but Dell Computer Corp and Hewlett-Packard Co are also quite prominent. Dell machines are widely held to provide very good value for money. Hewlett only recently began to shake its reputation for high prices, earned during the 80386 and 80486 generations; now that its machines are more affordable, a lot of customers are willing to pay what they view as a fair price for HP quality. Because all Intel Corp servers adhere to the standards that evolved from IBM’s Personal Computer (and particularly AT) designs, they are reasonably interchangeable. Whether they run Windows NT, OS/2, Santa Cruz Unix, Solaris or even MS-DOS and NetWare, iAPX-86 servers are functionally indistinguishable. They do vary in speed, expandability, reliability and serviceability, but the leading brands (and quite a few less prominent labels) are suitable for corporate settings in which the cost of an unreliable server can quickly out-run the machine’s acquisition price. IBM should be a natural leader in this market, and it was in the 80386 generation. IBM PS/2 servers may have cost more than alternatives, but there was no doubt in the minds of buyers that they were worth it. Unfortunately, IBM never took advantage of the Micro Channel bus in its servers. The IBM computers that use the asymmetrical multiprocessing that Micro Channel could have brought to the personal computer business are AS/400s, which use their own bus, not a Micro Channel. Micro Channel appears in IBM’s RS/6000s, but very smart device controllers, like the ones in AS/400s, aren’t available for the RS/6000. Go figure. During the 80486 generation, IBM held its lead, but not by much. Compaq offered EISA bus servers that had the speed of the Micro Channel and the less costly add-in boards that became the true industry standard for servers before the PCI bus appeared. EISA made it to the top for two reasons.
By Hesh Wiener
One, it was upwards-compatible with the AT bus; and two, because IBM mistakenly thought it could extract substantial royalties from other personal computer makers unable to otherwise build PS/2-compatible computers. Wrong! IBM had another chance during the Pentium generation, which is now nearly over in the server segment. All Big Blue had to do was give better value for money (which does not necessarily mean a lower purchase price) and corporate buyers would have opened their hearts and, more importantly, their wallets. But IBM has not offered a markedly superior Pentium server. Besides, how could IBM expect to sell computers when it kept telling prospective customers things like Trust me. Buy OS/2. That slash 2 concoction is apparently an abbreviation for the phrase run away while you can, but only
a privileged few at IBM seem to know it. Fortunately for IBM, the folks who named the database software DB2 rather than DB/2 knew this, with the result that IBM, in contrast to its experience with slash 2 products, has made billions of dollars and pleased thousands of customers. IBM can count itself fortunate that its board of directors chose a fellow named Lou Gerstner Jr, as chairman rather than one called Lou Gerstner/2. If the current trend continues into the P6 generation – and there is no reason for it to do so unless IBM hasn’t learned anything lately, which it hasn’t – the Big Blue share of iAPX-86 servers at mainframe sites will drop to something like 20% or 25% of the base. Compaq might get more than its current 40%. Hewlett and Dell could split another 25% of the market and the rest of the vendors will get scraps. Windows NT will be the dominant operating system in the P6 base, but the battle for the other crucial kind of software, database management systems, may not be won by Microsoft Corp. The strongest signal that Microsoft will emerge as the top database vendor is the nervousness of Oracle. We don’t know exactly what Oracle is worried about, but anything that gets Larry Ellison to settle his stomach with a diet of Apricots must scare the pits out of every other software house. Our prognostications about the P6 generation stem from the belief we expressed right at the start of this essay, that large enterprises are clever shoppers. And from what we see going on in the base, they are getting even smarter about the server business all the time. By Unix standards, NT is dirt cheap. By RISC standards, Intel servers are bargains. By any standards, the RAID arrays and tape back-up systems for P5 and P6 machines are very good values. Also, it makes technical and business sense to insist on software harmony among servers and the PCs they directly support. The PC operating systems corporations use are almost invariably one of the versions of Windows, for Windows is where you find the best applications at the lowest prices.
Adding to its own misery, IBM has managed to alienate the one marketing force that has worked on a volunteer basis as far as IBM is concerned, the third party lessors. While a finance company doesn’t usually express a direct opinion about a client’s choice of hardware, it does set lease rates based on its vision of the equipment’s future. Independent lessors are now predisposed toward disdain for IBM equipment just as they used to err on the upside. All it took was IBM Credit Corp eating the mainframe processor and RAMAC disk base and then sitting down to gobble up all the AS/400 financing in the world. Lessors no longer have much appetite for anything IBM has left on the table, which these days means PC servers and not much else. More accurately, they are scared and they would rather finance a Compaq server, even though that means competing with IBM Credit, which also seems to think a lot of Compaq iron. From the February 1996 edition of Infoperspectives, published by Technology News of America, 110 Greene St, Suite 1101, NY NY 10012, phone 212- 3349750 (C) 1996 Technology News of America Co, Inc. All rights reserved.