Original URL: http://www.theregister.co.uk/2009/09/10/oracle_sun_sparc_commitment/

Oracle and Sun taunt IBM with Sparcs

Throwing down another gauntlet

By Timothy Prickett Morgan

Posted in Servers, 10th September 2009 15:21 GMT

If Oracle keeps saying that it wants to be in the server and storage hardware business long enough, maybe some people will start believing it.

Ahead of its Oracle OpenWorld customer event, which is being held from October 11 through 15 in San Francisco, the company has been running advertisements that poke a stick at IBM, which is shaping up to be Oracle's biggest and most important IT rival if the $7.4bn (net about $5bn) acquisition of Sun Microsystems is approved by the antitrust authorities in Europe.

Two weeks ago, as El Reg already reported, Oracle said that it was determined to show that Sparc-based servers could best Power-based servers running Big Blue's own DB2 database on the TPC-C online transaction processing benchmark. A test, by the way, that Sun has shunned since 2001 since its iron did not scale as far as IBM Power boxes or Hewlett-Packard Itanium machines and, more importantly, since Sun had been charging way too much for its UltraSparc-III and UltraSparc-IV servers.

But Oracle threw down the right hand gauntlet at IBM's feet, suggesting that it would show on October 14 that Sun iron plus Oracle's databases would be able to deliver many millions of transactions per minute (TPM for short, and two digits of performance at that) compared to the 6 million TPM that IBM's top-end 64-core Power 595 (using 5 GHz Power6 chips) currently delivers.

It is a fair guess that Oracle and Sun will try to demonstrate that a cluster of quad-socket Sun Fire 5440 servers, using Sun's latest 1.6 GHz Sparc T2+ processors can deliver a lot more oomph.

By my back-of-the-envelope calculations, ten T5440s could deliver maybe 16 million TPM in a single rack without clustering overhead (that's ten separate images of the TPC-C test), and maybe 12 million or so once clustering for a single system image and other overhead is taken into account. It would also be very interesting to see Oracle and Sun to trot out the F5100 Flash Array, which El Reg has discovered is in the works. The F5100 is a 1U container that packs 80 flash modules and that sports four 3 Gbit/sec SAS channels to link it directly to servers, delivering 4 TB of capacity and over 1 million I/O operations per second. Such a machine would let a T5440 server just scream on the TPC-C test. Maybe enough to knock down the number of servers required, in fact. (My estimates above assume a fairly large number of 15K RPM disk drives per T5440.)

Today, Oracle and Sun are throwing down the left gauntlet at IBM's feet with yet another ad:

Oracle Sun IBM Second Gauntlet

This one even quotes Oracle chief executive officer, Larry Ellison, as you can see. And presumably - at least as far as Oracle is concerned - lays to rest any nagging issues Sun's customers might have about Oracle commitment to a hardware business that it has never really played in before. To be fair, Oracle has played hardware vendors against each other brilliantly, but that is not the same thing as bending metal and baking chips.

For what Oracle is outlining in this advertisement to be true, Sun had to have already nuked its Sparc and Solaris development efforts starting last year and continuing through the period when Sun started shopping itself around to IBM, Oracle, and presumably also Hewlett-Packard and Fujitsu. But we have it on paper now: Oracle is going to spend more money on Sparc development and Solaris development and is going to more than double the number of people selling and servicing Sparc iron.

No one ever doubted that tightly integrating Oracle with Solaris could be an important differentiator in the server racket, and Oracle will no doubt not only tune its database for Sparc/Solaris iron, but also middleware and application software. This is why you bother to spend billions on a hardware company: to create an integrated and easier to support platform.

But still, Oracle's promises laid out above lack specifics, the kinds of details that customers who are being asked to invest in Sparc T servers or Sparc Enterprise machines (which are based on Fujitsu's Sparc64 chips and the Japanese company's server designs). Where are the chip roadmaps and the server roadmaps? Where are the accompanying Solaris and storage roadmaps? Any merger is a sales pitch, not an ultimatum, after all, right?

As contentious as the acquisition of Compaq by Hewlett-Packard was back in 2001, with big shareholders and Hewlett and Packard family members opposed to the deal spearheaded by controversial HP president and chief executive officer, Carly Fiorina, HP did the right thing and laid out its product integration plans to show customers, partners, and government antitrust authorities what it would do with Compaq - which had already eaten Tandem Computers and Digital Equipment and which had complex product lines and issues of its own. You can argue about whether HP's plans were smart or not, but at least you knew what they were.

Not so with Oracle and Sun since the acquisition was announced in April. Both companies have talked in vagaries and have let rumors about product lines being killed off fester for months on end without comment from either Sun or Oracle top brass. Sun's executives - not just those at the top, but those in charge of key product lines - have been muzzled when they could be used to talk up the future products. And deal with dead products, such as Sun's once-and-not-future "Rock" UltraSparc-RK processors and their "Supernova" family of servers, for instance, which are widely believed to be dead.

The uncertainty surrounding Sun's servers has predictably had a chilling effect on its server sales. (Billings for Sparc Enterprise machines fell by 54 percent in the fourth quarter ended in June, Sparc T machines were up a tiny bit, and x64 server sales dropped by 9 percent.) With the EU regulators taking their sweet time mulling the database implications of the Oracle-Sun deal, Sun is going to have one or more bumpy quarters for server sales, but Oracle could smooth the road ahead out a little by telling customers what it will and will not do with Sun's products - with a lot of detail - should the deal go through.

Which it almost certainly will, particularly if Oracle lets go of MySQL as part of the deal. Oracle can always re-launch a MySQL product at some future date anyway, thanks to it being open source. ®