Damage limitation is the name of the game at IBM Corp after it shot itself in both feet with the announcement of its latest round of cuts in which it made little effort to conceal the fact that it recognises that the days of the mainframe are numbered. President Jack Kuehler and mainframe chief Nick Donofrio called in the Wall Street Journal to discuss the future of the mainframe as IBM sees it – and the medium term future is a parallel agglomeration of hundreds – and later thousands – of what sound as if they are the basic microprocessor that powers the ES/9000 9221. As they described it, the idea is to create an eight-way, or even 16-way node of these things, and then wire scores of these nodes together in a parallel configuration.
Genuiuses
In its first incarnation, promised for this year, this will be a machine with about 100 processors, designed specifically to accelerate database searches. In the medium term, the idea, as we reported in CI No 2,046, is to create a parallel version of MVS that will enable IBM to offer a general purpose parallel mainframe; our information in the earlier story appears to have been incorrect in that it now appears that the building block processor is not a RISC. The parallel mainframe will be in addition to the RS/6000 RISC-based parallel machines due next year, which will start out with models designed for scientific supercomputing before evolving into commercially-oriented models as well. At the same time as all this is going on, IBM says it will continue to develop its conventional mainframes, making them cheaper and more powerful, and that these will include increasingly cheap parts. But it acknowledges that the mainframe development team is no longer regarded as the company’s biggest asset, simply one of several equal children in the large systems group. All of which seems to underline the extent to which IBM has lost touch with reality: perhaps, to be charitable, the company is run exclusively by geniuses. The same issue of the Wall Street Journal that outlined IBM’s mainframe plans included the latest in its regular Manager’s Journal feature, this one written by Mortimer Feinberg, chairman of a Manhattan consulting group. The piece was headlined Why Smart People Do Dumb Things, and inter alia notes that Smart people tend to surround themselves with other smart people and suggests that when people of extraordinary brilliance form the palace guard of a company, the combined power of their intelligence can form an irresiistible force, pushing the enterprise toward the pinnacle – or the precipice. One danger is their unwillingness to admit the need for change – when smart people all agree with each other about a plan, they’re apt to stay with the plan too long, even after others have seen that the direction is wrong. The piece highlights another problem, the inability of bright people to listen to what those of lesser intelligence are saying. Itseems to describe what has been happening at IBM all too well – after all it was back in October 1984 that we wrote a piece headlined IBM Unprepared to Meet the Threat from Unix.
Back teeth
And what IBM’s response to all the Death of the Mainframe press it has received after its latest round of planned cuts makes clear is that the company can’t bring itself to acknowledge that most users are fed up to the back teeth with IBM mainframes and all their works – or more specifically with MVS. In most mainframe shops there are three kinds of application: those that can at present only be run on the mainframe, those that can be migrated over time to a cheaper open system, and completely new ones, which can be developed from the ground up on an open system. Given that mainframes are intrinsically expensive to run – the software costs an arm and a leg, and an army of people is needed to feed and water them – the majority of mainframe users want to invest as little as possible more in their machines. They will likely need more storage capacity, so new and more efficient disk drives will still be welcomed. But for most users, where m
ore processing power is required, they can meet the need with the cheapest used machine they can find: doesn’t matter that it’s not the latest technology: IBM itself used to boast that the 308X machines had a 30-year design life. Of course there are still a significant number of users for whom even a ES/9021-900 is not big enough, but once those users start seeing their siblings limiting new investment in IBM mainframe technology and diverting it elsewhere, they are unlikely to be so overburdened with intelligence that they fail to recognise that if the mainframe base is no longer generating the wherewithal, then IBM will not have the resources to develop the even bigger machines that they need, and they too will start looking around for alternatives.
Undying loyalty
IBM’s biggest problem is that it simply hasn’t treated its mainframe customers sufficiently kindly to retain their undying loyalty to the company, so that once mainframe users do start looking for alternatives, many of them will look outside IBM altogether, confident that even the most ignorant chairman has read enough of the headlines to be one jump ahead of them and be more than ready to back a move out of the IBM camp. And that is why people are talking about mainframe meltdown, and why the fingers of Jack Kuehler and Nick Donofrio in the dyke will do nothing to stem the gathering torrent. And the fact that it was the inferno in the mainframe shop that the company rushed to fight only draws attention to the fact that it has not yet woken up to the fact that the AS/400 base is tinder dry and that sparks are flying ominously all over it. The AS/400 is the one unarguable jewel left in IBM’s product crown, but all the good intentions with which it was launched have run out of steam. By some accounts, a full half of the System/36 base remains unconverted, and most people believe that even IBM has given up hope of converting it. Top end AS/400 users are running out of steam again, and users are beginning to experience that funny, familiar System/38 feeling all over again. Instead of talking pie in the sky about parallel MVS, IBM should have wheeled out the AS/400 specialists to reveal a new commitment of cash and resources earmarked to shoot the AS/400 into the performance and ease-of-use stratosphere – and about how IBN was finally going to make an AS/400 offer to System/36 users they really couldn’t refuse.
