When questioned about the reliability issues of using Intel and Linux -- as opposed to an IBM mainframe -- Ellison said: "A lot of people criticise and say that Linux is not as reliable as a big old IBM mainframe -- I happen to think it is, but it doesn't matter because if one, two or even three machines in the cluster fail, you still have more than 20 left -- so who cares? Nobody will notice. This provides complete fault tolerance." He also commented on the improved efficiencies of using clusters: "The old approach to reliability was to have a mainframe running your application, and if anything went wrong, you would switch over to another identical mainframe which is otherwise doing nothing. This is called hot-standby and was state-of-the-art in reliable computing. But in a cluster, the whole grid is working all the time, so it is dramatically cheaper." "One of the most shocking things about moving to the Dell grid, is if you want a lot higher performance, you are going to have to be willing to spend less. Linux and Dell will be the dominant combination in the enterprise," added Ellison. "It's a good start," said Weiss, "But we are still a number of years away from being able to measure the overall success of this venture. Although users do want the better price and performance characteristics of a standard platform, there have large investments and commitments in legacy infrastructures." But Weiss warned that there are still concerns regarding latency issues -- in terms of memory to memory transfer between near and far servers. The new Dell-Oracle clusters will start at $18,000, Dell said. That price includes two dual-processor Dell PowerEdge 2650 servers with direct-attach SCSI storage, the Oracle software licence and two years of gold-level support, which includes 24-hour, four-hour response-time service, Dell said. Munir Kotadia reported from New York. CNET News.com's John G. Spooner contributed to this report.





