Mainframe users should not hesitate to take advantage of the attractive CMOS option in every setting where the technology provides immediate benefits and an assurance of flexibility in the future. IBM has dramatically broadened the scope of its S/390 product line, encompassing the vast majority of legacy customers. It has also indicated that the impressive progress it has made to date will continue apace. Customers can plan on the availability of CMOS systems with engines of 60, or more likely, 80 to 100 MIPS by the end of next year. In addition, the new storage subsystem that will become available to users of 2003 processors during the first quarter of 1997 may be the basis of data libraries for users of other platforms, including but not limited to all S/390s. But the new IBM mainframes will not immediately serve customers with the largest systems because the 9672’s single engine speed of about 43 MIPS is not capable of carrying the workloads run by many users of H5s, H5 compatibles and Hitachi Skylines. The large market for IBM’s new CMOS systems will bring IBM record MIPS shipments during the next year and, possibly but not certainly, an upturn in mainframe revenue. It will also result in substantial increases in the MIPS at many sites, giving mainframe user companies the freedom to expand existing workloads and add new applications.

Hitachi faces memory lock-out

When Hitachi Ltd decided it would be better off buying CMOS S/390 chip sets from IBM Corp, its engineers were determined to build a better system. One consequence is that Hitachi’s main memory is not the same as that in the IBM’s 9672 or 2003 systems that use the same processor technology. Hitachi’s mainframes may or may not run better than IBM’s because of this unique memory, but we are confident Hitachi’s finances will run more slowly. If Hitachi had not only been CPU-compatible with IBM, but also memory-compatible, it would have been able to gain access to two important market segments: original memory and upgrade memory. Nearly all the IBM CMOS machines not owned by IBM through its captive finance subsidiaries are controlled by Comdisco Inc. There is no question that Comdisco would be willing to finance IBM frames with Hitachi memory, if the option were available and, if Hitachi memory cost less, customers would be receptive. But the market for IBM-compatible memory is far broa der than its potential in a fraction of new processors. One of the most striking characteristics of the CMOS mainframe user base is its willingness to sign leases that will end up lasting five years or more. During the terms of these leases, customers will buy more MIPS and when they do they will need more memory. In addition, IBM’s Multiprise 2000 line uses main memory (and an S/390 CPU chip set) to control internal disks. If IBM succeeds in its attempt to sell single-frame S/390s, there is bound to be a lively market in memory upgrades. Finally, to the extent that IBM persuades its CMOS base to use their machines to provide Internet or intranet services, customers are going to find that they will install unexpectedly large amoun ts of main memory so their servers can compete in the GIF-tossing Olympics, at which IBM hopes to be somewhat more successful than in last summer’s similarly named sport competition.

(C) 1996 Technology News of America Co. All rights reserved.