It is becoming abundantly apparent that it was IBM Corp and not Hitachi Ltd that was the loser in IBM’s decision not to take the Japanese company’s Skyline monolithic ECL mainframe OEM when such a deal was on offer nearly three years ago (CI No 2,418). As we forecast at the time, IBM would not lose an enormous number of customers by trying to railroad its entire mainframe base to CMOS processors and Parallel Sysplex, simply that most of the ones that would dig their heels in would be IBM’s biggest and most important customers. So it has proved. Hitachi’s Skyline has won customers ranging from AT&T Corp through AMR Corp to Lloyds Bank Plc, WalMart Stores Inc, L M Ericsson Telefon AB and Dow Chemical because those customers did not believe that the Parallel Sysplex cat’s cradle would adequately meet their requirements, and were not prepared to go to the effort and expense of breaking their workloads up into smaller sub-applications. Meta Group Inc reckons that Hitachi Data Systems Inc’s share of the mainframe market soared to 22% from 7% last year, and its sales for the year, due next month, are forecast to exceed $2bn, compared with $1.6bn in the year to March 1996. Moreover, IBM has been hopelessly caught out by the Year 2000 problem and the threat of a single currency in Europe: large users have to parallelize their biggest applications to run them on Parallel Sysplex configurations, and IBM reportedly has to concede that only about 100 CMOS users have so far done so. After all, who wants to waste the 12 months or more Gartner Group Inc is quoted as saying it takes to parallelize an applications workload when they have the Millennium and monetary union to contend with? According to Information Week, IBM is due to announce a 60 MIPS CMOS uniprocessor by mid-year, but the 120 MIPS Skyline CPU is expected to go to between 160 and 180 MIPS by year-end. For some strange reason, Hitachi proudly insists that the Skyline is full of CMOS as well as having ECL for the speed-critical circuitry, despite the fact that for those in the know, when it comes to performance, ECL is a tiger and CMOS is a wimp.
