The edge that IBM holds over its plug-compatible competitors has been shrinking all through the 1980s, and with it the premium that users can justify paying for the IBM badge and blue cabinets rather than red ones. The mathematics become even more problematical if the short life for the high-end ES/9000s posited by consultant William Husband have to be factored in. Janice McGinn got the message.
Hitachi Ltd’s marketing strategy includes several new releases in 1991 and 1992, and 1993 will see the introduction of a further four machines ranging from two to eight processors and offering between 240 MIPS and 500 MIPS. So postulates William Husband of the Meridian Group in a wide ranging analysis of the IBM and plug compatible marketplace over the next few years (CI No 1,544). Apart from the EX 310 and EX 420 which are due in the first half of 1991 (CI No 1,441), the EX 110 and EX 210 will be available in the third quarter of next year, and they will deliver 44 MIPS and 80 MIPS respectively. Husband suggests that Hitachi will launch an EX 620 in the fourth quarter of 1991 and the six processor machine will execute 225 MIPS, to be followed in 1992 by an eight processor 255 MIPS EX820. Husband says that the EX 820 will be the last of the current EX series, which dovetails with his theories about IBM launching MVS/ESB with a 38-bit address register in 1992 and a new architecture by 1996. He believes that the 1993 releases will effectively double Hitachi’s MIPS rating and performance. The 310A, 420A, 620A and 820A will come in at 240 MIPS, 300 MIPS, 400 MIPS and 500 MIPS respectively.
Teradata
Husband believes that Hitachi is collaborating with Teradata Corp on a back-end database to compete with IBM’s DB2, and he reckons it will be available in 1991. On Hitachi’s strategy for the existing EX range, Husband forecasts that the EX 10 to EX 44 – 5 MIPS to 24 MIPS – will be available only on request since they cannot compete with the ES/9000 range, and users should look for discounts of up to 30%. The EX 50 to EX 100 – 23 MIPS to 87 MIPS – will either be cut across the board or individual discounts of up to 50% may be negotiated. Hitachi Data Systems, 80% owned by Hitachi and one of Hitachi’s two European distributors, declined to comment on Husband’s scenario except to say that it wouldn’t reveal its marketing plans to anyone outside the company, and certainly not to the Meridian Group, and neither Hitachi Data Systems nor Teradata are prepared to confirm that they are collaborating on a back-end database. Hitachi Data may have reservations about Mr Husband and his crystal-ball gazing capabilities, but he has few such reservations about the merits of plug-compatible manufacturers.
He says that if one compares the plug-compatibles with IBM in terms of cost when IBM is not offering discounts or volume purchase amendments, then IBM comes in a poor second. In terms of CPU life – range of memory sizes and longer product cycles – and environmental costs – floor space and power – the plug compatibles, especially Amdahl, are way ahead. Husband insists that if users are not getting a discount from IBM, then they are paying too much, and the company is already discounting the ES/9000s by as much as 35%. On incremental upgrades, Husband claims that going from a 3090-200 to a 3090-300S has a cumulative cost of $9.6m. But moving from an Amdahl 5890-200 to a 400E will cost around $7.7m, almost $2m less than the IBM equivalent. Amdahl also comes in a winner with CPU market life, and Husband says that while IBM and Hitachi offerings run for around 20 months, between 1982 and 1988, Amdahl equipment had a market life of 35 months. He acknowledges that Amdahl’s advantage is likely to diminish after 1992 in the high-end arena, but he thinks it will continue for machines under 200 MIPS. He says that when IBM did not make the 3090s available for rental, it was a reflection of the shortening product cycle, and he believes that will continue until IBM’s new machines are introduced in the mid -1990s, ushering in a longer cycle. Husband is forecasting
that the next few years will witness much faster uniprocessor performance, and he reckons that Amdahl and Hitachi are going to outstrip IBM by around 50%. By 1994, Summit will execute about 70 MIPS, while the other two could be touching the 150 MIPS mark.
16-processor
The number of tightly coupled processors is set to grow rapidly, and Husband claims that both the plug compatibles and IBM will be offering 16-processor machines by 1995, although he suspects IBM will lag in 1994 when Hitachi and Amdahl introduce their respective offerings. Husband argues that while IBM machines have strong, if declining, residual values, the cost of buying IBM kit is much higher than either Amdahl or Hitachi. Nonetheless, Hitachi machines still have very poor residual values, and that must be remedied if the company is to be regarded by users as the equal of Amdahl and IBM. In terms of reliability, Husband rates all three as excellent, and on current and future memory requirements, says that both Hitachi and Amdahl have 2Gb offerings while IBM continues its incremental phasing approach. Many will question Husband’s analysis of the IBM and plug-compatible market and much of what he says is controversial and speculative. However, few would disagree with his observation that while compatible technology implementation lags IBM, innovative implementation is led by the rest of the pack.
