On competing with IBM—a subject Ellison has analyzed in various ways in recent months, including by describing how he wants to emulate the IBM of 50 years ago—Ellison's most intriguing comments came as he talked about Oracle is now planning to push its highly successful Exadata system beyond data warehousing and into the broad field of high-end transaction processing. (That belief, by the way, is one that's not shared by IBM, and we'll get to that in a moment.)
"Exadata is the best example of this clustered architecture where we have collections of machines, storage, and networks all built together with the software tuned for the hardware. What we've been able to do with Exadata—and there will be other machines like or clusters like Exadata coming out for middleware applications, pretty much for all the software that we sell," Ellison said.
"Exadata Version 2 handles not only data warehouses but also handles transaction processing--we added a lot of flash memory into it—and we're gonna have new Exadata models—I don't want to preannounce anything, but we're gonna get better and better at very large-scale transaction processing. Our intent is that the Exadata line will challenge the biggest IBM P Series machines and beat them badly in performance, reliability, and cost. And we think we can do that: in transaction processing, we're twice as fast, and in data warehousing, we're 10 times as fast. Those are our goals: to be twice as fast as IBM's biggest, best box and again, at a dramatically lower price."
As you might expect, IBM said that Ellison's argument sucks eggs, via this comment from a company spokesman who said that if Oracle intends to thrust Exadata into various markets, then by definition it is anything but an optimized system:
"There is a fundamental philosophical difference between Oracle and IBM," the IBM spokesman said. "Oracle is trying to convince the market that one system can efficiently handle both transactional and analytical workloads. IBM believes that systems should be optimized for particular workloads.
"For transactional use cases, Oracle sells Exadata. For analytical use cases, Oracle sells Exadata.
"On the other hand, IBM recommends systems that are optimized for particular workloads. For transactional use cases, IBM recommends DB2 pureScale, which is a tightly integrated, pre-configured, and pre-optimized hardware, software, and storage stack," IBM said.
"For analytical use cases, IBM recommends the IBM Smart Analytics System, which is a tightly integrated, pre-configured, and pre-optimized hardware, software, and storage stack that includes ETL, data warehouse, Cognos reporting, SPSS analytics, and much more."
Industry analyst Joe Clabby of Clabby Analytics added that while it's one thing for Ellison to claim that he intends to beat IBM's P series, it'll be quite another thing to actually do it.
"Sparc is so far behind P series that Oracle-Sun would have to make incredible investments over time to try to catch up—I'm not talking about a couple of hundred million dollars here, I'm talking many multiples of that," Clabby said.
"So as for using Sparc in these big transactional systems, here's the question: yes, you could do it, but do you want to do it? Nobody that I talk to is buying UltraSparc anymore—they're bailing out on it like crazy."
In spite of Clabby's staunch skepticism, he did admit to a glimmer of believe that Ellison can in fact compete effectively against IBM.
"I can't believe Ellison spent that kinda money ($7.4 billion) for that kinda hardware. But if anybody can do it—if anybody can make this work—it would be Ellison. He's very impressive."
Bob Evans is senior VP and director of
To find out more about Bob Evans, please visit his page.
For more Global CIO perspectives, check out Global CIO,
or write to Bob at [email protected].