Later this year, we expect, IBM will provide details of its most important development for large systems in several years: a database co-processor. IBM might even be able to put the machine into volume production during 1993 but may settle for a system to offload CICS terminal handling to start with. But either way, its offering of a database engine will have an enormous and far-reaching impact. There will be great debate surrounding the engine and many questions will be raised that cannot be answered except in terms of likelihood. So the first and possibly best result of the expected IBM announcement will be a renewed interest in the future of large systems. Pundits who have insisted – with the great assurance that is a constant companion of shallow minds – that the demise of the mainframe is imminent, will be forced to do a bit of dancing. The co-processor using groups of small engines, they will explain, is not a mainframe at all and in fact evidence that the mainframe is a gone gosling. The simple truth is companies with large central data processing facilities will still need these unpleasant, draughty, noisy rooms. An issue that will soon become far more pressing than arguments over corporate real estate is the future of large systems database management programs. Some of the best software technology comes from independents who must be very clever indeed to withstand IBM’s DB2 and IMS. The widespread assumption about IBM’s forthcoming engine – and one to which we subscribe – is that it will be a machine designed to support one software scheme only. That software will be DB2 with SQL capability (or, if you prefer to think of it in other terms, SQL with an accessible DB2 engine). It is too early to say whether IBM’s competitors in the database arena will be able to adapt their products to the new hardware at an affordable cost.

By Hesh Wiener

A retreat by these vendors to the market defined by systems smaller than those likely to get the engine – to 9121s, for instance – would be a temporary solution at best. If IBM has any sense (which is in itself a topic worthy of serious consideration) it will offer versions of its engine for the entire ES/9000 range; should it do so, the 9221 version might be nothing larger than a circuit board or two that fits in a standard rack. Ultimately, if IBM can offer a range of efficient CICS platforms and database engines, the applications processor(s) in a large system might become autonomous machines with a technological life of their own. An example of this has been evolving at IBM for many years with the AS/400. While the computer systems may reside in a single cabinet (plus some outboard boxes for disk drives), within the AS/400 are terminal support units based on fast general-purpose microprocessors, disk and tape controllers using comparable technology, an intelligent memory subsystem, one or more applications processing engines and a DBMS that owes its power and economy to hardware and microcode. It is not too soon for mainframe customers to start thinking about the new course on which IBM will soon embark. Although there may not be terrible penalties for taking a cautious approach to new and untested technology, it would be imprudent to make excessive investments in the hardware and software that is now offered by IBM and its ilk.

This essay appears in the May 1993 edition of Infoperspectives International, published by Technology News Ltd, 110 Gloucester Avenue, London NW1 (C) 1993 Technology News Ltd.