Pluto dedicated mainframe processors, revived Fort Knox AS/400 9370 in IBM’s future Although not as well organised as it would have us believe, IBM is better organised than it looks, according to International Technology Group chief, Brian Jeffery. In contrast to other commentators, Jeffery also believes IBM is regaining some of the financial ground it lost during the early 1980s, and predicts a 12% to 15% revenue growth in the 1990s. Kicking off this week’s Xephon-hosted IBM Futures conference, Jeffery argued that the company is currently embroiled in a transition phase, dating back to research and development work initiated in the 1970s. Since 1986, he added, IBM has been exploiting the research, equipping products with evolutionary platforms designed for technical optimisation throughout the 1990s. Short-term, Jeffery anticipates that a number of 1970s-based developments will also lead to the introduction of new across-the-range features between 1991 and 1992. Chief manifestations will be the arrival of the Summit mainframe, an AS/400-9370 hybrid, RISC-based workstations, a new PS/2 architecture, and changes within the networking field. Addressing each product area in turn, Jeffery pinpointed removing the multi-processing performance degradation – the drop in performance which occurs when additional processors are attached to mainframes – as IBM’s chief preoccupation. He believes that IBM will achieve this by splitting up the arrangement of processors on the bus, and that, once accomplished, CPU expansion potential will be unlimited. Summit will also provide a significant improvement in channel rates, although tangible price-performance benefits will not become a reality for at least three years. By the mid-1990s, Jeffery also believes that an even more powerful type of mainframe system – codenamed Pluto will have arrived. This will form the core of IBM’s long-term vision, a raised floor central mainframe complex, to which function-specific modules – transaction processing, database and communications processors – are attached via fibre optic cabling. Early examples of this kind of complex have already been spotted at the University of Michigan, and IBM’s Montpellier site. Meanwhile remote sites, manned by AS/400s and 9370s, will link in to the central mainframe over Token Ring networks. IBM’s overall plan, concluded Jeffery, is distributed, but firmly mainframe based computing. The message for users, he added, is enormous gains in CPU power; however, those who want mainframe-based computing will have to wait at least five years for its delivery, while those who don’t should start examining alternatives for the future. Turning to networking, Jeffery argued that IBM is finally beginning to show some kind of commitment to OSI Open Systems Interconnection. However, in his opinion, the OSI subsystem products introduced last year may set up an alternative which can be run in conjunction with IBM’s SNA, but do not provide an integrated OSI offering. Confirming repeated gossip that all is not well in the AS/400 camp, Jeffery also predicted that large System/36 will move to the AS/400-9370 hybrid, when it arrives. Meanwhile, Jeffery claims that System/36 migration to AS/400 is simply not happening, adding that IBM’s Application Business Systems division in the US has failed to meet AS/400 targets during the first half of this year. Turning to the PS/2, Jeffery argued that IBM never designed it as a personal computer. Over the next year, Jeffery believes IBM will bring out a 80486-based machine specifically designed for co processing, complete with a new architecture and MCA-based functionality improvements, to provide seamless integration with the mainframe. Customer delight Meanwhile, he added, IBM will effectively relegate stand-alone, non-Micro Channel, machines to a schoolhome-based offering, turning them into 3270 dumb terminals, which they have been all along. He also looks for a minimum of 24Mb memory on a PS/2 in two to three years, and believes that when integrated with a host, the host-to-workstation p
rocessing load will be split 40%:60% respectively. Eventually, Jeffrey envisages two different styles of co-operative processing. The two level scenario involves PS/2-to-mainframe links over Token Ring, while the three level alternative replaces Token Ring with an AS/400 or 9370. In the US, he says, big chunks of IBM users have indicated that they will opt for the two level approach. Finally, Jeffery turned to IBM’s marketing approach, saying the goal of customer satisfaction had now been transcended by customer delight. According to Jeffery, IBM is now offering its customers all the support they can eat. It is also looking to co-ordinate everything it sells – particularly third party software – via a single, account marketing interface. As always, the message to users is mixed, and the motivations ambiguous. At one level, the interface is good news for users, and a serious IBM attempt to build relationships with large accounts. At another, Jeffery warns that the move may prove an effective means of regaining control of accounts, after the erosion IBM suffered during the 1980s. – Sophie Hanscombe DFSMS makes systems programmer redundant Bob Blakeman of Integrated Computer Systems and Cybernetics managed to upset one or two delegates by suggesting that system managed storage has numbered the days of the systems programmer and the technocrats had best cast their eyes around for other fields of endeavour. According to Blakeman, System Managed Storage will give the user of the 1990s total freedom of information management. He went on to address current user fears that IBM’s storage management plans will leave them with little autonomy. The user or owner of data will have to establish ground rules and group data in relation to performance levels, workload patterns, available storage types and the cost of storage. The ultimate objective would be to create as few pools and as little input-output as possible. The user will control active data and define which pools to use for different types of data, but the placement of data is to be controlled by DFSMS. Blakeman sees active and medium storage as typically disk and long term storage as tape. All this is to be made possible by VSAM which is device independent and facilitates properly managed SMS. The 3990 controller was appraised and found wanting by Blakeman, largely because of slow disks. Nonetheless, he looked forward to the real thing which should have multi-gigabyte capacity, fully integrated solid state disk, automated library management and fibre optic connection. On the future of disk he sees a smaller form factor; rack mountability; improved connectivity with possibly eight way pathing; higher availability and data portability; use of a thin film medium as well as thin film heads, and actuator level buffering. Blakeman sees a switch from storage management to memory management and left potential System Managed Storage users with a five point plan. Determine which strategy you want, determine techniques, acquire the software, run a pilot scheme, and research before implementation – the costs are high, but so are the benefits. – Janice McGinn
Content from our partners
