Only weeks after Amdahl Corp announced that its six and eight-way 5995M systems have slipped to the fourth quarter, the company’s Dublin manufacturing plant is shipping its first 5995 Model 4550Ms. The US announcement was perceived as welcome news to IBM Corp, which has still to announce an eight-way system. Analysts and commentators believe that day is not far off, and that Amdahl’s problems are providing IBM with both breathing space and the opportunity to dent Amdahl’s top-gun position. Nonetheless Alan Bell, marketing director of Amdahl Europe, is maintaining the company line and insists that IBM is playing catch-up and struggling like mad to get an eight-way out into the marketplace. Bell may be right, but he’s fighting an isolated rear-guard action by maintaining that IBM is having enormous difficulties. IBM has not announced an eight-way, but few industry-watchers believe that it is struggling with the technology. 390 architecture is designed to facilitate such a system, and IBM’s decision to hold fire is undoubtedly a marketing one. IBM is capable of delivering an eight-way.
Whether it does so, and when, are not technology issues. If it does announce – possibly just as Amdahl ships – then Amdahl has lost competitive advantage at the very time the existence of glasshouse technology is coming under scrutiny. It is possible that Amdahl is due to ship its best and most sophisticated technology to date. It is equally possible, probable even, that the mainframe world is about shrink and become very expensive niche technology. Amdahl acknowledges that consolidation could decimate the number of MVS sites, but it claims that the demand for large mainframes and powerful single images is growing. Even if consolidation runs at 40% over the next few years, David Thomas, large systems marketing manager, reckons that 900 out of 9,000 mainframes will be Amdahl systems. But given the potential size, availability and price-performance, he says he would rather to operate in that scenario than the one that exists today. Thomas is of the view that mainframes will be optimised to handle both compute-intensive applications and those already engineered for MVS, which he says are estimated to be worth some $700,000m. Future imaging and storage requirements need enormous horsepower, and they, along with increasing consolidation, the need for Sysplex technology and the importance of application development, will be primary mainframe drivers. That fits very neatly with Amdahl’s strategy. Not only is the company leading the pack in terms of power and talking of 1,000 MIPS systems acting as superservers, but it claims that the Huron application development system provides the capability to interoperate, interface and integrate.
That is, code and applications go where they are best suited whether it be superserver, workstation or personal computer. So says Amdahl. But sceptics remain unconvinced. Yes, imaging and storage will be technology drivers, but will they they be the drivers of mainframe technology? It could be argued that increasingly sophisticated data compression techniques actually reduce the need for the mainframe. And as regards application development, Amdahl has a version of Huron for 80486-based workstations and Unix-compatibles due to to ship at the end of this year. The company may have led the way with Unix on the mainframe, but with the exception of AT&T Co, UTS (to be honest) has had limited success. Alan Bell denies this, saying that some users are running UTS native, and that mainframe Unix has been a tough ride simply because there are a lot of MVS bigots out there. Nonetheless, IBM validated Amdahl by its AIX announcements, and Amdahl is now finding itself bidding UTS against AIX. Given the move towards distributed computing and the decision by Barclays Bank Plc to use Unix at the branch level, shouldn’t Amdahl also be moving in that direction? No, says Bell, control is crucial. But, despite Amdahl’s protestations, some people are more interested in its plans for a RISC-based wor
kstation than its attempts to bolster the disintegrating world of the mainframe. Ken Gorf, director of open systems, has conceded that the Key Computer Labs’ Unix supercomputer was not right for Amdahl’s market (CI No 1,680). Subsequently, Amdahl transferred staff from the Key project to its 7300 line of Unix-only mainframes from Fujitsu, but Bell acknowledges now that the 7300 was very much a toe-in-water experiment that didn’t come to anything very much, largely because of poor price performance. So, 18 months down the road, Amdahl’s RISC-based development remains a mystery. Bell acknowledges that SPARC development based on Sun Microsystems Inc’s chip is ongoing, and that Amdahl will introduce a high-end client-server driver product. But he refuses to say when a RISC-based processor with an Amdahl badge will hit the marketplace.
Perhaps coyness is the nature of the beast, but while Amdahl is prepared to give installation and trial figures, it remains tight-lipped on naming users. This is especially true of Huron, and while the Inland Revenue will present a Huron paper at the UK Computer Measurement Group meeting in Brighton this year, Bell refuses to be drawn on other large users. Excluding the UK, there are five paid-up sites in Europe with another 12 on trial, and a further 11 in the UK, four of which have paid for the thing. It costs anything from $350,000 to $1m, plus a monthly usage charge. Bell says he would like to be able to name names, but persuading blue-chip organisations and the likes of Marks & Spencer Plc to advertise Huron – which is still a largely untried and unknown quantity, commercially speaking – is not easy. Amdahl can leverage its existing relationships with the top 50 IBM system users to push Huron – either stand-alone or as a bundle – but each installation demands the same quality of service and support as a hardware installation. Consequently, while Amdahl is keen to see Huron prosper, it is targeting specific customers rather adopting the hard sell. Which is a logical approach since Huron cannot afford to have any bad references. But it’s a Catch 22. Huron cannot afford bad references, but if convinced users won’t endorse the system, it seems to create a credibility gap. It begs the question of why having made the commitment and paid for it – only the Inland Revenue out of nine European customers is prepared to stand up and say that it works. Amdahl may claim that Huron develops applications up to nine times faster than other means, but the world at large has yet to see the thing working, or at least to hear those claims coming from someone other than Amdahl’s European marketing manager.