Peter Junker, vice-president of Computer Corp of America, claims that CCA is the number two software company in Japan, second only to Software AG. He also says that Japan is one country where IBM has to operate like a plug-compatible manufacturer and have microcode to support the Japanese operating system. Junker forecasts that within the next few years, IBM is going to have to face a Japanese invasion of Europe and the US, and Big Blue is fighting a losing wicket. He believes that if Hitachi and Fujitsu launch a Japanese operating system on a worldwide basis, then plug-compatibles can ditch the performance-hindering microcode, and kill IBM on its own territory. Which sets the stage for a repetition of the Japanese scenario. Of the 6,000 IBM-like mainframes in Japan, Fujitsu and Hitachi each has around 2,500, and IBM comes a poor third. If Junker is correct and the Japanese are going invade Europe and the US, then Computer Corp is well positioned to capitalise on their likely success. Junker says that CCA is the only vendor to offer Kanji support, and able to implement software on the Japanese hardware. Junker’s musing on the future of IBM was prompted by the launch of Horizon with the Model 204 database system for IBM’s new baby 9370 Models 10, 12, and 14. IBM actually supplied CCA with a machine nine months prior to the launch. A chummy thing to do, and it illustrates the love-hate relationship that exists between IBM and Computer Corp. However, Junker couldn’t help crowing over a recent set query benchmark conducted by the University of Massachusetts and audited by Coopers and Lybrand. The benchmark trial was developed by Professor O’Neil of the University as an alternative to the debit/credit benchmark which has been criticised for being too simple to measure general commercial applications. O’Neil’s version takes set queries from numerous table rows at the same time. The benchmarking was done on an IBM Model 4381-3 running MVS/XA, and suggests major differences in performance between the two databases. For example Model 204 elapsed time was lower than that of DB2 in 64 of the 69 queries and these varied between 0.3% and 79% of the corresponding DB2 query. This means that Model 204’s response time outperforms the DB2 by 8 to 1. Other figures include a 16 to 1 better CPU elapsed time and a 26 to 1 cost advantage in favour of Model 204. Junker insisted that he was not suprised by the results although he admitted that IBM’s status as the industry leader meant that other people might be. He is understandably chuffed with the benchmarks, and confident enough to claim that DB2 will never come close to Model 204 in large database and query applications.
