Quietly - very quietly – Oracle has been dropping out of the HPC market. We’re finally seeing some outward reaction to the company's internal moves with news stories (The Reg here and HPCwire here) discussing Oracle’s retreat from HPC. No comments from Oracle either confirm or deny the move, of course. I’ve been hearing …
What about Coherence ?
Oracle didn't buy Tangosol that long ago, and Coherence is actually one of the better offerings they have now. It would be a shame to see them ruin yet another product.
The move by Oracle does seem odd, particularly when Microsoft are beginning to move heavily into the area. I'm not entirely sure Oracle are really happy when extending beyond the enterprise database space.
Oracle is a sales/marketing company. Sorry DBA fans, but its true, products work just well enough, but are by no means leaders in their respective technology areas.
So HPC is a loss leader. As the author points out, HPC R&D leads towards understanding how to build things like clouds, virtualization, high performance networking, all of the systems kit that we'll see out over the next 5 years.
Is this a bad thing for Oracle? Probably not. Oracle has always wanted to stay away from hardware but in order to get Sun's 'Software Crown Jewels', they had to purchase the whole company.
Oracle should, like IBM, stick their toe in to the grid/cloud computing. This is important for Oracle's future because the grid/cloud allows them to add in a higher end analytics to their portfolio. Unlike IBM which released a brain dead fork of Hadoop. (32 bit Linux with the only differentiation being a 'configuration tool'? C'mon)
Oracle could buy Cloudera, or license Cloudera to perform L3 support until Oracle develops the technology in-house.
So yeah. Oracle is missing the boat because they are still thinking in 2D space when you're fighting a 3D war. ;-) (You need to see an Old Star Trek episode to grok that...) ;-)
It was a movie, actually.
The post is required, and must contain letters.
HPC is the future
Today's "high performance" is tomorrow's mainstream. Dropping out of HPC won't hurt immediately but they'll miss out on long term innovation.
9 gets you 10...
"But the firm has brought out some innovative hardware along with some design approaches that separated it from the pack, at least to a degree. I’m not sure exactly what will happen to that gear over time;"
I can hazard a guess; it's going to stop innovating, and sit on the innovations/services acquired, and wait until it can find a deep pockets defendant against whom it can act as an aggrieved plaintiff in a patent case.
Sun didn't really die until Oracle acquired it; now Oracle's busy killing off the sum of the parts. Look out OpenOffice, MySQL; you're next on the line for commercialization.
1. Oracle 11 has the title 11g where g stands for grid, so they do follow buzzwords a bit. Expect v12 to 12c, where c==cloud.
2. No idea about oracle and cloudera, but I do know that the Hadoop team, recommend Linux only in production machines, Sun JVM. Because it is the only thing that other people (facebook, yahoo!) test at scale. Some patches for solaris have gone in, but other JVMs are very much a "you deal with the problem yourself" kind of issue.
3. if you look at the future of Hadoop-style HPC, it's commodity hardware with commodity SATA storage and commodity gigabit ethernet. Oracle could play here, but they have to want to.
(disclaimer: Hadoop committer stevel, ASF member).
I hate to say it, but Larry is right
HPC at Sun drove technology in the wrong direction. Wildcat diverted million of dollars and who-knows how many man-hours. HPC requirements added to the cost and delay of UltraSPARC-III and Starcat/Serengeti. HPC complexity helped make Millennium unbuildable.
If HPC innovations trickled down to products that actually...y'know...generate revenue, that would be one thing. But historically, that has not been the case. One of the things that killed Sun was a this tolerance for staring at shiny objects, instead of watching the bottom line.
Leave HPC to the spooks and the grad students.
Sad for us, right for Oracle
Much as it pains me to say it, this sort of laser-like focus on profit is why Oracle are buying Sun, and not the other way around. Oracle don't need to keep a bunch of HPC wizards in house, they have deep enough pockets to watch how the market matures and then co-opt or buy the new guys out. So if Supa-Dupa-Connect becomes the new I/O interconnect standard, they'll just support that, or buy it in, and they're done. No need to throw money at a bunch of projects only a handful of which will see the light of day.
I suspect that the Reg readership skews strongly to a sentimental feeling that Sun was a really really cool company with some truly ace talent. Which it was, for years. Unfortunately they pissed it all away and now we have Oracle making those hard decisions for them. Ugly, but there it is.
1, 2, out
This is the old Jack Welch (former CEO and Chairman of General Electric) axim - 1, 2, out. For a particular market, you have to have immediate designs on being number one or number two. Third place is not a strategy.
Oracle wasn't going to be number two in HPC any time soon. So they decided to cut their losses. Makes sense if you subscribe to the Welch-ian management style.
If HPC was really a good investment ...
SGI would be the world's leader in computing, Convex would have bought HP, not the other way around, and DEC's Alpha would have ruled the processor world.
While it is correct to see the HPC influence on current Oracle products (Exadata is basically an InfiniBand cluster of x86 computers, combined with optimized processing nodes), Oracle does not need to be in HPC to do that R&D (they will do it anyway), and it is stupid to try to monetize R&D over the money losing HPC business (see SGI).
- Updated Zucker punched: Google gobbles Facebook-wooed Titan Aerospace
- Elon Musk's LEAKY THRUSTER gas stalls Space Station supply run
- Windows 8.1, which you probably haven't upgraded to yet, ALREADY OBSOLETE
- FOUR DAYS: That's how long it took to crack Galaxy S5 fingerscanner
- Did a date calculation bug just cost hard-up Co-op Bank £110m?