The financial case for a mainframe is based not only on hardware costs, environmentals and maintenance. It is also very much affected by the cost of system and application software. Yet IBM Corp has no direct control over application costs. Most mainframe applications have been written by customers. Even where users have bought applications, the suppliers are independents, not IBM. IBM’s control over the cost of mainframe software, then, is confined to the systems programs that shape the S/390 environment. System software constitutes the foundation on which customers erect suites of applications. It is also the basis of competition in the processor business, because IBM’s systems software makes old and new S/390s equal and creates an opportunity for other manufacturers to offer IBM-compatible central processors. Just as IBM has been under a lot of pressure to reduce the price of its mainframe hardware (and related running costs), it has also suffered when customers perceive its systems software as too costly for a particular job. In response, IBM has made its mainframe systems software more affordable. But the assumptions about customers that underlie IBM’s software pricing have not been given the same careful review that IBM seems to have applied to hardware prices.

Financially prudent

Although there is a logic to IBM’s practice of reducing system software charges per MIPS of processor power as users move to bigger computers, it is not the only course of action open to IBM. IBM could try to level its pricing structure. It could even decide to charge more for big users than small ones. We are raising this issue because IBM is currently most vulnerable at the very top and very bottom of the large systems segment. The high end issue stems from IBM’s decision to cease making large engine machines until it can create their central processors using CMOS technology. IBM is losing business to Hitachi Ltd Skyline systems and will continue to do so until it can demonstrate that its Parallel Sysplex clustering scheme is both technically sound and financially prudent. IBM did institute premium software pricing for machines with more power than the largest 9021 under a fee scheme it calls IMLC. This may have affected some customers’ plans, but big software, like big iron, seems to be selling briskly these days. The immediate situation is not hard proof that Skylines and high-priced software for them will sell well indefinitely, but it does mean that plenty of customers who examine the total cost of large mainframes have decided the machines are priced at acceptable levels. In any event, big shops can examine all IBM’s pricing schemes, choose the one that seems most appropriate and then bargain for discounts on software that often run 25% to 30% of list price but in special cases reach levels of 35% or greater. At the other end of the S/390 line, IBM faces very different problems. IBM’s competition in the mid-range comes only in part from its own used machines and not at all from IBM-compatible processors. Instead, users whose machines have under 25 MIPS compare mainframes to Unix and NT systems or to AS/400s.

The results are apparent in the market statistics. Few if any new users are coming into the S/390 fold at the low end, while thousands of companies are installing servers that provide alternative software environments. Software pricing is only one of the issues causing the S/390 base to diminish in population. There is also the matter of software functionality and standards. In the Unix or NT world, mid-range users have access to a wealth of applications programs that are not available to S/390 shops. IBM has enhanced MVS to provide Unix features and, in theory, this enables Unix applications to be converted for the mainframe. But there does not seem to be a great deal of such activity. Software vendors are barely able to keep up with the demands of competition in the Unix and NT markets; they have little interest in chasing customers who are unlikely to

buy applications even if they are available… if history is any guide. IBM charges quite a bit for system software in the S/390 mid-range and users find it affordable mainly if they are using lots of legacy applications. They stay with their established systems because the hardware and software are very reliable and, in each customer’s domain, very well understood. But unless IBM does something to make the S/390 dramatically more attractive for small and mid-size workloads, its base will decline slowly and then, when it is quite a bit smaller than it is today, rapidly disappear. This process is under way in the System/36 base, which happily persists for now. However, it could be gobbled up by NT or Unix at any time. IBM’s greatest challenge is to maintain its software base in the high mid-range and at the top end as it perfects the Parallel Sysplex technology that promises to give the S/390 a new lease on life. Along the way, IBM is going to have to make some big changes. It must not merely provide hooks for Unix applications and TCP/IP communications; it must excel. It must not become arrogant about the superior reliability of its operating systems: alternatives will only grow better while IBM’s own software, undergoing renovation, could go through a period of difficulty. Before long, IBM must make its mainframes not merely open in the sense of accepting function calls from Unix programs, but rather open in the sense of fitting into the attractive new world of Internet- intranet technologies. This will not be easy. And it may turn out to be a practical impossibility.

Force upgrades

The S/390 world is a 31-bit universe, not even as capable of directly addressing large memory spaces as the most powerful personal computers, let alone Unix or NT servers. IBM might well conclude that MVS, VM and VSE could absorb much more in development costs than they are worth… particularly if future generations of CMOS mainframes will require a 64- or 128-bit architecture to provide the performance characteristics mainframe customers require. IBM could offer versions of Java generators, Netscape servers (or other Web server programs) and numerous other features that are in demand without destroying the well- regarded frameworks of its strategic operating systems. But do mainframe customers really want these features if they force upgrades that automatically turn into higher software costs? So far, there is no universal answer to this question. But IBM is aware that it is on the minds of most if not all of its mainframe users, customers who have already installed innumerable Unix and NT servers to surround their mainframes. The most promising answers may come as a surprise to IBM, and to many big customers, too.

By Hesh Wiener

From Infoperspectives, May 1996, (C) 1996 Technology News of America. All rights reserved.