Storage strategies at mainframe shops are being reviewed. Managers of the largest IBM Corp mainframe installations, briefed on the forthcoming Seascape family of storage products in the past several months, find themselves in the lurch. Seascape will not appear in 1996 as planned. Instead, IBM will offer a third generation of Ramac arrays next year and beef up its 3990 disk controller. Customer reactions vary from resignation to open scepticism regarding IBM’s future in mainframe storage. Competitors rejoice at IBM’s woes and, not surprisingly, are making sure every prospective disk buyer is aware of the confusion within IBM’s disk storage products group. Seascape is a storage subsystem based on a scalable controller and various disk and tape devices optimised to work with the new controller technology. The 3590 Magstar tape was to be the first product in this family, and its delivery to mainframe users has slipped by nearly two years. It is possible, but not likely, that IBM will be forced to abandon the Seascape project entirely and instead buy or build alternatives. Periodically, unsubstantiated rumours that IBM will acquire Storage Technology Corp surface. While StorageTek does not have a unified storage scheme to rival Seascape, it does control the automated tape library business and its line of mainframe disk subsystems is, technologically at least, well ahead of anything IBM can deliver in the near term.
Bit of mirth
Gossips have also suggested that EMC Corp might buy StorageTek, although financial and managerial aspects of the hypothetical merger make it very unlikely. Those who speculate about a marriage of EMC and StorageTek do so with a bit of mirth, pointing out that IBM’s reaction to the creation of a mainframe storage rival that would be both wealthier and more effective than Big Blue itself is undeniably an intriguing notion, however improbable. There are several issues raised by the deferral of Seascape that have immediate bearing on mainframe customers: the presence of three generations of Ramac disk equipment is likely to shorten the economic life of not only 3390 disks but also Ramac I and Ramac II machinery. Similarly, delays in deployment of Seascape will add to the lifetimes of other disk arrays, including those offered by EMC, StorageTek and Hitachi Ltd. Delays in Seascape combined with competitive pressures may force IBM to make its next generation of storage subsystems more open. IBM may have to consider the convergence of 390, AS/400 and AIX storage. IBM’s market share in large systems storage, now in the vicinity of 35% and slightly below that of EMC, could plunge below 25% during 1996 as users seek faster storage devices than IBM can provide. All disk prices could drop rapidly as IBM finds itself unable to compete on any basis except price in the large systems disk segment. There are still a lot more unknowns than facts, but the likely course of events does not appear to be very good for IBM, or for users who have chosen IBM 3990/Ramac storage. Fast disk subsystems can raise mainframe throughput very considerably. Where jobs use data that can be properly staged in disk controllers, IBM’s offerings are likely to perform poorly compared with those of its rivals. The result is a decrease in the effective throughput of large systems and consequently an increase in the cost of information processing. The actual effect of disk speed on computing varies from site to site and from job mix to job mix, but there is one measure of disk value that transcends the performance claims of all the vendors: price. These days, IBM may be the price leader in disks, but this is a dubious achievement. Users are willing to pay a premium for faster EMC, StorageTek or Hitachi equipment. IBM may have to come up with some greatly improved marketing plans to counter the growing power and cleverness of its competitors. All these discouraging trends could reverse in two years, when IBM has new storage subsystems and much faster CMOS mainframe engines. IBM still controls mainframe operating systems, which gives it enormous power to shape the directions taken by the high end of the market. But it could find its influence confined to central processors not whole mainframe systems.
