Oracle Slapped Over Anti-Power Advertising Campaigns
August 6, 2012 Timothy Prickett Morgan
Larry Ellison, co-founder and CEO of Oracle, has never been shy and he likes to compete. The company also likes to keep its marketing messages as simple as they are aggressive, and contrary to the do-nothing marketing approach by IBM, Oracle has something to prove since it acquired Sun Microsystems more than two years ago. Still, sometimes its claims about how it is better than one competitor or the other cross the line.
Twice in the past year, in fact, according to rulings from the Better Business Bureau‘s National Advertising Division, the advertising part of the North America organization business practices watchdog.
Back in April, the NAD reacted to a challenge from IBM about an Oracle advertisement from last fall that claimed the Sparc SuperCluster T4-4 clustered system “runs Oracle and Java twice as fast as IBM’s fastest computer,” which the advertising identifies as the IBM “P795” server, adding that the Sparc SuperCluster T4-4 cost $1.2 million, while IBM’s Power 795, as the machine is actually called, costs $4.5 million. This ad ran on the front page of the Wall Street Journal as well as a full fold-out page in the interior of the paper. I never saw the ad myself, but I certainly saw and reported on the performance claims Oracle made for the machine.
In his presentation, Ellison called the Power 795 “one big, expensive, single point of failure,” which is funny, but he did not get into how the comparison between a Power 795 backed by DS8700 disk arrays from IBM was rigged to look bad against eight racks of Exadata X2-2 database clusters. If you read what I wrote, I did just that, also pointing out that the comparison that Oracle made did not include the cost of the systems software, which was much cheaper on the IBM box and, at list price anyway, utterly dwarfed the cost of the hardware. And the funny thing is, IBM never did issue any kind of counter claims or pick apart the comparisons or make better ones.
How is it that I do IBM’s job? Can I get IBM to pay me for it?
The Exadata machine is jacked up on special Exadata storage arrays, with hybrid columnar compression on the database to speed up certain kinds of transactions, as well as having massive amounts of data compression. The machine is also packed with more flash memory to speed up I/O and therefore the database. A fair comparison would have used a flash-based PureScale cluster, or at least admitted that it was comparing two different kinds of systems–a database cluster and a big wonking SMP box. The NAD didn’t say jack about that, and quite likely because it doesn’t understand what either IBM or Oracle are talking about. What Ellison said back in September 2007 was that an Exadata cluster spanning two racks could do transactions anywhere from 10 to 50 times faster than the Power 795 backed by four DS8700 arrays. He also said an Exadata setup spanning eight racks delivering three times the processor cores, 3.5 times the storage capacity, 4.5 times the flash, and has 13.5 times the storage bandwidth, and 10 times the storage I/O operations per second. Such an eight-rack Exadata setup would cost $8.8 million compared to $18.6 million for the IBM setup.
Anyway, after the March complaint by IBM, Oracle pulled the ads.
Oracle has been running another set of ads, which claim Exadata runs 20X faster than IBM Power machines, and that this data came from a European retailer who switched platforms. IBM complained to the NAD once again, and correctly claimed that the 20X comparison between Exadata and Power was “overly broad,” which it sure as heck is. Oracle said it was a case study, not a general claim, and said further that we would all get that here in IT Land. The NAD disagreed, and on July 26 handed down a ruling against Oracle and said that Ellison & Co had done the right thing and stopped distributing the disputed ad materials.
Oracle plans to appeal the NAD’s finding, apparently. It seems more likely that Oracle will announce a bunch of new systems at Oracle OpenWorld in late September and early October, and make all kinds of claims then.
What IBM needs to do is stop messing around and show off all the goodies in its servers, storage arrays, operating systems, and database software and actually prove that its stack is as good as Oracle’s “engineered systems.” IBM spends far too much time talking about Smarter Planet and not enough time talking about Smarter Systems. And relying on the NAD to slap Oracle’s wrist (if it is even that) after the message is out there in the market doesn’t really fix anything.
And another thing: I have been asking IBM for nearly a decade, and probably since 1995 if my memory serves, to build a cluster of low-cost, entry machines to compete with Oracle’s gridded database engines (that’s what the “g” in Oracle 11g stands for, after all) head-to-head. And I have also suggested that IBM should build those clusters from OS/400 and now IBM i systems. Take the fight to Oracle. Stop reacting and act.