Posted by: Ed Scannell
Exadata Database Machine, IBM, multi-core servers, Oracle, Sun
The good thing about Sun canceling development of its 16-core Rock processor means Oracle now has one less set of multi-core servers to fret over with its cores vs. processor licensing policy. The bad news is the new Oracle, hoping to compete against the likes of IBM and Hewlett-Packard from chips to business process software will come to battle with a few less bullets.
I am only half kidding, of course, about the good news part. For better or worse (worse many users say) Oracle doesn’t fret much over whether it should license servers by the processor core or by the box. It unflinchingly continues to license by the number of cores, which could prove an expensive proposition for some buyers.
The bad news I may not be kidding about so much. With Sun kicking its five-year old Rock project to the curb, Oracle can’t make the next leap in the game of performance leapfrog it plays with chip rivals IBM and Intel.
In fact, Sun has failed to leap a couple of times having canceled its UltraSparc-V chip project earlier this decade because it essentially ran out of development funds. The company rushed out its UltraSparc-IV chip as a stop-gap product that didn’t do much. And when Sun was skipping a leap, a couple of other times it was very slow to leap, being months even years late in delivering a new chip.
Sun officials have talked consistently the past few years about the Rock being a game changer. The chip was designed to achieve a much higher per-thread and floating point performance, along with greater Symmetrical Multi-Processing capabilities than its Niagra family of chips.
Rock, which was to anchor the company’s Supernova line of servers, was supposed to really shine when it came to handling high-end data facing workloads including database servers. Just guessing here, but I think a machine delivering great database performance would be important to Oracle.
With Rock out of the game Sun will continue to use Fujitsu processors, which will be fine, but hardly represents the game changer the company was hoping for. One positive aspect to all this is that Sun’s research and development costs just got significantly lower. This won’t make Larry Ellison unhappy as he tries to complete the $7.4 billion acquisition of Sun.
But once again Sun has opened the door wide open for a number of competitors to rush through, most notably IBM. With one less competitor at the high end, IBM figures to rip away more server market share from Sun over the short term. Some speculated over the last day or so that Oracle might start to emphasize the next gen Niagra III Sparc chips, as well as pushing Solaris on x64 servers fueled by Intel’s upcoming Nehalem EX servers.
Sun has been relying on chips from Fujitsu for its larger servers while it waited for the Rock development to be finished. Now it is likely to just continue using Fujitsu chips, which should lower research and development costs.
But what does the lack of a “game-changing” chip like Rock do to Oracle’s plans to sell vertically integrated hardware-software stacks (as has been rumored) ala its Exadata Database Machine? It could hurt its newly acquired hardware business for sure, but perhaps more importantly hold back its flagship database business at the high end in some key markets, along with other proprietary and open source software offerings.
We knew Oracle was buying some damaged goods with its acquisition of Sun, but I am not sure if Redwood Shores was assuming its chip business might be this damaged.