At a recent UK executive briefing on mainframe disk subsystems staged by Meta Group, IBM guru and large systems analyst Marc Butlein intended to talk all morning about storage directions. But over coffee he found that the attendees wanted to hear about something else: what the hell is happening to the IBM Corporation, and how will its problems affect its large system customers? So he prefaced his talk with 45 minutes on the state of IBM today. First, he set the grim scene for IBM: average hardware selling prices (note that the term list prices has disappeared) are falling fast; competitors have never been more successful; 3390 disks and ES/9000 processors are losing market share; bureaucratic costs and overheads are too high; the ES/9000 Summit models do not offer enough added functions to encourage users to upgrade; there is little understanding of the software business within the company; and the sales force has become relatively ineffective. He noted that central processors, where IBM still has a strong position and makes good margins, are becoming less and less important, but mainframes nevertheless have dominated IBM’s development strategies. Mainframe economics are a now big problem for seller and buyer alike.
Too high
For IBM, the cost of selling a mainframe is over 30% – much too high. For the customer, prices are so high that many are now finding new mainframes unaffordable. Clearly, IBM must change, but it isn’t easy. IBM is a victim of its success. The main problem for IBM is that its mainframes have been too profitable. Even with all the discounting, IBM’s gross profit on mainframes is around 60%. Although IBM does not want to lose this money by dropping prices, many customers, says Butlein, are having serious affordability problems. The result of all this? Demand for mainframe power is flat, and users are concerned about cost pressure and system complexity. Downsizing and outsourcing is accelerating, there is increasing commoditisation of hardware and vendor loyalty (mainly to IBM) is dead. There are, Butlein said, three questions that users usually wanted an answer to: first, how will IBM pricing change? Second, How will IBM charging structures change? And what role will competitors play? Butlein suggests several developments, but every scenario includes continuing instability and problems for IBM over the next few years. Underlying all the problems for IBM and its mainframe customers is one big assumption: that large system demand is not affected by price.
By Andrew Lawrence
But Butlein argues that users are increasingly affected by prices, and that they are making up shortfalls by leasing (mortgaging the future), by buying in the second user market, by buying plug-compatible machines and by downsizing. IBM is starting to react to this, and recent trends in pricing will continue: users can expect IBM (and rivals) to accelerate price-performance curves of mainframes, increasingly to bundle products together, to revise its charging methods, to attempt to reduce the cost of ownership, and to charge more for software and less for hardware. Users will increasingly be asked to sign three- to five-year fixed price (unified pricing) deals. Having set the scene, Butlein then moved onto storage. IBM mainframe users need to start looking at alternatives to IBM’s 3390-type disk subsystems, he says. He says they are not just expensive, they are unaffordable today and intolerable tomorrow. Butlein asserts that IBM and its immediate competitors (Amdahl Corp, Hitachi Ltd, Storage Technology Corp) are deliberating keeping prices high for 3390-type storage devices. Had it been a competitive market, the price would be one quarter what it is today. But, says Butlein, change is on the way and disk centricity is eroding. Users will buy any storage technology that suits their application, in the form of low-cost disk arrays, mass storage devices and optical disks. In future, storage will be much more important than it is today, and the storage controllers will be the most important devices. The disk controller
is already becoming a growing determinant in disk selection, and a growing proportion of the costs. The ability of controllers to increase channel utilisation, to employ Fast Write features and to improve performance through large caches are all key. Butlein is expecting a range of important products over the next two to three years, starting inevitably with Storage Technology’s Iceberg RAID Redundant Array of Inexpensive Disks device (CI No 1,850). Among the new products from IBM will be: a universal disk controller, with fault-tolerant design, 3990 and 3390 emulation and 9Mbytes-to-17M-bytes per second data transfer.
Escon only
It will use Escon only and will be a key part of Sysplex: it will ship in 1993. IBM will add Shared Expanded Storage, which can be shared by up to 64 central processing units. It will be a Sysplex integration vehicle, with storage costs of less than $500 a Megabyte. There will be a new 10.8 disk device, the 3390-4. It will probably the last in the 3390 family. It will be have 4Gb per head-disk assembly and will be announced and shipping this year. A disk array will be launched by IBM eventually, says Butlein, but it is in no rush to endorse StorageTek’s Iceberg. It will probably not be shipping before 1995. And the much-predicted archival tape library to rival StorageTek’s Nearline will not be shipped until 1993, says Butlein. Optical disk will be shipped in 1993 and offer pricing per Megabyte at 10% of disk prices. All of which sounds dilatory in face of the growing competitive threat.
