AT&T Corp would not comment on the notion that it is one of Intl Corp’s two other potential OEM customers for the scalable parallel processing line, but suggested it is always interested in the possibility of sourcing technology rather than building it, and pointed to the rich heritage of iAPX-86 technology and culture it has amassed. The key differentiator for AT&T is its highly-tested Unix System V.4 MP-RAS environment set against the untried, untested Opus. Its second concern is that Intel, IBM Corp and other parallel architectures are built on generic bus topologies, not tailored for decision support and data warehousing as is its Ynet system, designed for the Teradata back-end massively parallel database system. However, AT&T has slipped up badly with a much-delayed BYnet bus successor and non-existent iAPX-86-based System 3700 massively parallel processing machine it had planned. Of late, AT&T has positioned its aspirations for the next-generation bus as separate from system concerns. It says it has the bus up and running internally, and details are planned for this summer. System implementation details have been pushed far out from the bus itself. We underestimated what it would take to develop the bus to the next generation architecture, AT&T says, claiming to have a fully functional technology in testing. AT&T, which denies having received and rejected offers to buy the Teradata technology, will unveil a promised Unix version of the Teradata database running on its 3500 symmetric processing systems at DB/Expo this week.
