Windows NT will soon be available for IBM mainframes: company executives have been telling outsiders that an announcement is imminent – apparently on the processor itself, in contrast to the AS/400, where it runs on an 80486 co-processor. IBM has not been shy about licensing other software from Microsoft – its desktops typically come with Windows, its Pentium and PowerPC servers with Windows NT. Putting NT in the sealed box of a mainframe is a new and unexpected development, although the concept was mentioned as long ago as November 1994 (CI No 2,551). Early indications are that the target market for the offering is the 9672 base. These sites are among IBM’s most progressive main-frame users and they certainly are likely to have a strong interest in developing applications based on such popular server programs as Netscape. The other programs in the NT world that are commonly used when NT servers provide the primary support for workgroups are unlikely to be of great interest to mainframe users. The most obvious benefit to IBM would be to reduce the number of users that would otherwise attach outboard NT servers to 9672s or other main- frames. Open Once such servers are installed, migration of applications from the central system to the server becomes a likelihood that IBM knows will lead to a reduction in the strategic importance of the central mainframe and, eventually, to the possible replacement of the mainframe by alternative systems not made by IBM. A less specific but potentially powerful byproduct of NT support would be vigorous reinforcement of IBM’s claim that the S/390 is an open platform that permits customers to select with great freedom the most appropriate software for every application, even if that software originated far afield from the IBM mainframe world. If main-frame users become concerned about the long-term viability of their mainframes they will not only move applications to other systems; they will also move data. IBM certainly does not want that to happen. A third benefit to IBM is that the addition of NT would open new markets. The S/390 – more specifically, the 9672 – could be positioned as an all-in-one computer. Unix system vendors sell their machines on the same basis, but users know that in practice most substantial Unix servers are configured with outboard workgroup servers. This provides considerable price-performance benefits, but at the cost of more complex technical support. By offering a simpler solution to medium and large businesses, IBM should be able to outflank HP and Sun, particularly at sites where technical expertise is scarce. IBM could also use the availability of NT to simplify migration of the current S/390 to a 64- or 128-bit architecture. The porting of NT to a new hardware platform is costly but still not expensive compared with moving MVS or other IBM mainframe environments. More importantly, it could be done much more quickly. Then, with NT as a bridge supporting end users and handling some of the services that are forcing IBM to consider a major extension of the S/390 architecture, delays in the transition of environments that must continue to support large collections of legacy programs would be masked. Further, customers could be promised that they could have the best of the legacy and Internet-intranet worlds – which is the result they are not obtaining by encapsulating their mainframes in networks of external Unix or NT. IBM is taking a few chances by offering NT to its mainframe customers. But they might not be as dangerous as they first appear. One issue that will be watched by Wall Street analysts and IBM shareholders is the possible impact of NT on IBM’s mainframe software revenue. The low cost of Windows NT might make IBM systems software pricing seem excessive and users might take advantage of the proximity of NT to accelerate their migration away from traditional operating environments. Mainframe users are generally aware of NT costs already, but the second issue is a very serious matter.
By Hesh Wiener
There is a strong case to be made for standardization of software and NT, like Unix, is a hotbed of packaged applications development. Some of the big players in the NT and Unix markets, such as Oracle, have also offered their programs to IBM mainframe users, but for a number of reasons the success of Oracle and its ilk in the mainframe world is not as great as one might guess, mainly because mainframe users are reluctant to ditch established custom applications. But the temptation of the Internet-intranet concept with standard browsers on desktops could be a catalyst for long overdue changes. IBM is ahead of the game in this matter, having said it would provide full browser support in its existing operating systems, but unless IBM puts Netscape server under MVS is might find itself trying to catch up in its own captive market. On the other hand, by tying browser support to the mainframe file systems, it could spark tremendous growth in the volume of data stored on its central machines, more than offsetting any shift of revenue from its own software division to Microsoft and other outsiders. We think IBM is acutely aware of the vulnerability it has to Microsoft. If NT on the mainframe becomes the dominant force shaping applications, NT on other platforms would certainly become more widely accepted. A prominent casualty would be Unix, the RS/6000 family and AIX. So it might turn out that IBM’s greatest problems would arise not directly in the mainframe market but indirectly in the RS/6000 market that sometimes appears to be a natural successor to the System/390 and at other times looks like a digression. The memory limitations of the current S/390 architecture are incompatible with a large body of applications for which NT is a popular choice. In a Parallel Sysplex, customers could conceivably separate their large engine needs from jobs that could be run on CMOS processors with smaller single engine capacities. But so far there is not much field experience to indicate how well this would work out. We do not believe there are more than a couple of dozen Parallel Sysplex systems used in routine production work at this time and it would be unwise of IBM to assume that it will be able to sell the Parallel Sysplex concept to a broad market during the next year or two. So far, a lot more mainframe customers have chosen to avoid or at least defer the Sysplex issue by ordering Skyline machines. One factor in IBM’s strategic thinking is the possibility that other companies might port NT to the mainframe even if IBM doesn’t. Specifically, HItachi Ltd is at least thinking about it and, like IBM, it seems to be of the opinion that NT is more appropriate for its forthcoming CMOS mainframes than for its Skyline behemoths. Hitachi Data Systems believes the presence of NT in the mainframe market would pressure IBM into reducing its systems software prices, and that might make it easier for Hitachi to sell its S/390-compatible processors. However, a change in IBM software pricing at the low end, now seriously overdue, might not be accompanied by a similar reduction in prices at the high end. The most that Hitachi could count on if it ported NT to the S/390 architecture is an increase in business due directly to the availability of NT itself. That may not be all it would like, but it seems to be a far better basis for a business plan than the theory that IBM could be pushed around in its own territory. We expect an announcement from IBM very soon, followed by a long period during which IBM will attempt to interpret the reaction of its customers. The pricing of NT will most likely be kept under wraps for as long as possible, as would the price of any special hardware required to bring NT under the skin of a 9672 or other eligible IBM processor. But in the end, NT on an S/390 will most likely cost about the same as it would on another system of comparable power.
From Infoperspectives, May 1996 Coyright 1996 Technology News of America Co Inc, 110 Greene Street, New York, NY 10012, 212 334 9750.
