Feeling lost in the fast-moving world of IBM? Here’s the state of play of the AS/400 as laid out by UK AS/400 Business Manager Malcolm Haines. The move to a RISC-based architecture for future AS/400s has been widely reported, but Haines says that the current BiCMOS processors will suffice for around another two years yet, […]
Feeling lost in the fast-moving world of IBM? Here’s the state of play of the AS/400 as laid out by UK AS/400 Business Manager Malcolm Haines. The move to a RISC-based architecture for future AS/400s has been widely reported, but Haines says that the current BiCMOS processors will suffice for around another two years yet, with the company pushing up the clock speed. At which point RISC will have to take over. At first glance that might look surprising since the AS/400 architecture is theoretically capable of running 14 processors and the current top of the range model F95 has only four. However Haines says that the processor count is unlikely to grow since the overheads become prohibitive. Perhaps more worringly, it is not yet clear whether the future generation of RISC-based AS/400s will be any more amenable to parallelisation.
In the meantime, the company is thinking about clustering AS/400s, perhaps via the new fibre adaptor,so that multiple machines can work co-operatively – although clustering is probably not a term that IBM will use, says Haines wryly. As for application portability between the CMOS and RISC versions, he points to the Integrated Language Environment, ILE, that was released as part of OS/400 version 2 release 3. A brand new ILE-compatible C compiler has been announced, though it is yet to ship and there are statement of directions regarding REXX and Cobol Compilers. ILE will be key in ensuring applications work on the new hardware, as well as providing performance enhancements, according to Haines. The Application Systems Line of Business used to work under political constraints that made sure that it didn’t tread too heavily on the mainframe business.
By Chris Rose
Those constraints have gone now, and yet Haines says for various reasons we should not expect to see future AS/400s challenging the top-end mainframe market, although it will become a stronger challenger around the low-end and middle of Enterprise System’s territory. Despite being a self-confessed AS/400 zealot, Haines is not blind to the product’s weaknesses, and readily recites a list of its shortcomings: not terribly good at compute-intensive tasks; a maximum disk capacity of 165.2Gb (effectively halved with the use of disk mirroring); a paucity of tools for managing large networks, and finally a lack of fault-tolerance. The last point is already under attack with the Rochester development team looking to see whether the multiprocessor AS/400s – such as the four-way F95 – can be made more robust. Currently if one of the processors fails, the machine collapses in a heap, but Haines says that work is in progress to let the others take over, more neatly.
When it comes to total processing power, however, the AS/400 will always be limited by the commitment to remain air-, rather than water-cooled – despite the fact that IBM’s rivals in the mainframe world seem to be able to deliver performance matching that of IBM with air-cooled machines, even if the air does have to be gale-force. Finally there is the question of the future merging of the RS/6000 and AS/400 CPUs, first reported in IBM System User (CI No 2,165). Haines obviously finds the idea of one machine for all, possibly based on Taligent, conceptually attractive, but says he has not heard anything in IBM about a merging.