Users of IBM 9021 mainframes don’t have to worry about their immediate situation. Their machines will continue to get support from IBM and third parties. The 9021s are among the best mainframes IBM has ever built: they are versatile, powerful and reliable. But 9021 users do have to be concerned about two things: the past and the future. The past is important because many of the strategic applications running on 9021s were developed long ago. In most cases, these programs have been modified almost beyond recognition over the years. User enterprises have treated these programs not as temporary artifacts related to a particular technological environment or a brief era in computing history. Rather, user organizations have come to view their strategic mainframe applications as permanent or semi-permanent assets.

Coal-fired boiler

But this is not the case, nor can it be in a world of rapid technological progress, fierce commercial competition and dynamic social change. The mainframe is no more permanent than the steam engine or the coal-fired boiler. For many years, it was the best industrial machine for large-scale information processing and, in many settings, it still bears that distinction. The anchor holding the 9021 in place is no longer a simple economic one. In most settings, the 9021’s longevity is more a confirmation of the strategic value of applications wedded to IBM’s durable large engine computing environment than an endorsement of the underlying hardware. The persistence of the 9021 is also a monument to failure, for IBM did not want its large engine mainframes to enjoy their present vital role this long. IBM had hoped to replace 9021s with 9672-E CMOS mainframes, and it is clear that it cannot. IBM may not be able to give the 9672-R the engine power of a 9021, not this year anyway, and probably not next year either. The 9021 will give way to other processors in a year or two or three. But the replacement for the 9021 will not be chosen by IBM, the way the replacement for the 3090 and the 308X were before. This time the successor to the biggest IBM computer will be picked by customers… and there may not be a single choice with the nearly universal appeal of the 9021. Some users – and we believe the majority of the current 9021 base – will migrate at a measured pace to 9672-R processors. But others will prefer the very large engine machines of the Hitachi Ltd Skyline series. Still others will recreate their strategic applications on systems that are exotic by current standards: IBM RS/6000 SP parallel processors, Hewlett-Packard Co servers, multiprocessor AS/400s or even super-servers based on large numbers of Intel Corp P6 chips.

By Hesh Wiener

The emerging strategy may not favor a single core processing system but instead a mix of machines chosen not so much for their hardware but because their software offers the best protection against a trauma like the one now being felt by 9021 users. Thus the next generation machine on which today’s mainframe users will build their strategic systems may not be hardware but instead a database management system combined with an Internet-style networking scheme. And there is no assurance that the software will come from IBM. Other suppliers whose software is available on a wide range of systems (but not necessarily on IBM mainframes) may currently have a sustainable lead. It is not too early for 9021 users to begin planning their next moves, even if their 9021s will remain in use for another couple of years. If strategic applications are to be put up on new systems – even ones based on the S/390 architecture – this may be the right time to initiate the development effort that will ultimately sever programs and databases from hardware. While few large users are keen on making their software and data collections more generic – it can be a costly, difficult and risky job – we do not believe that it is prudent for users to wait for an IBM solution that fully protects established software and data. IBM does not now have an acceptable answer

for users who require 9021s and we do not believe it will be able to offer an easy upgrade path in the near future, if ever. Our point is not that IBM will cease to be the pre-eminent systems supplier to large enterprises. We expect the opposite. But we are beginning to doubt the extensibility of the IBM S/390. Users whose jobs can be supported on engines of 20 MIPS or less can take the path IBM offers with its 9672 family. The 9672’s engine will be improved and if that design reaches some practical limits, IBM will be able to prop up the 390 framework on RISC or other chips capable of emulating a 30, 40, or 50 MIPS 390. But users whose growth is already hampered by the 9021’s 60 MIPS limit are not necessarily going to be able to jump to an IBM S/390 processor with equal or better effective single engine speed. There are practical limits to every computer architecture and if the S/390’s limit is now 20 MIPS in CMOS and 100 MIPS in Hitachi’s mix of bipolar and CMOS, it might not be wise for users to assume the computer makers will be able to deliver reasonably priced upgrades. IBM would be the first to point out that on a pure price-performance basis, its mid-range systems provide far better value for money than mainframes. Unfortunately, these systems do not provide the absolute power required by large enterprises. IBM similarly has shown that for a number of demanding applications its SP machines make very effective use of leading edge chip and interconnection technology.

Like to pay less

Moreover, outside the mainframe world IBM has done a very good job containing the cost of software, maintenance and environmentals. The facts suggest that IBM is not simply taking advantage of users’ inertia but rather that it is pricing its 390 products at levels the market will accept. The evidence lies in the excellent sales results IBM has reported and it is made no less compelling by the complaints of users who would like to pay less for their big systems. Nevertheless, users of 9021s must explore ways of stretching the useful lives of their mainframes while at the same time considering new directions. From the March 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.