BLOG UPDATE: After writing this piece back in December 2006, I found a page on searchdatacenter.com about HP’s promotion to get Sun SPARC customers over to Proliant server. Yes, that would be a no-brainer move, but not painless…enter Transitive. Along the same lines of what I stated in this blog entry, that news about HP included the following bit about Transitive:
In conjunction with the certification announcement, HP publicized a new relationship with Transitive Corp., which ports software across multiple processor and operating system pairs.
Transitive’s QuickTransit for Solaris/SPARC-to-Linux/x86-64 solution enables applications that have been compiled for the Solaris on SPARC to run on certified 64-bit HP ProLiant platforms running Linux without requiring any source code or binary changes, HP reported.
Oracle no longer has to port the database to such a wide array of platforms for the sake of garnering market share. They won that battle—rightly so. Who else had such an optimized RDBMS on so many platforms? Nobody, period. But most of those platforms are dead. So, imagine the reduced cost for Oracle to produce their database—someday in the future—on only one platform. Guess what platform that would be? Linux of course. “Oh boy, he’s fallen off the deep end”, you say. No, really, read on.
Porting Is Expensive—Beyond Labor
It costs Oracle to maintain multiple ports not withstanding that fact that the varying productlines supply engineering resources. Even Sequent had 27 engineers stationed on site at Redwood Shores. All that additional free manpower aside, the very existence of so many ports is a cost that is hard to explain to shareholders and analysts when it is an expense that Microsoft clearly doesn’t have to bear. Remember, the parties that matter the most at this point of the game are the analysts and shareholders.
I’m a huge fan of “real systems”. You know, the sorts of systems where life-and-death applications are running and the people managing them can sleep at night knowing their system isn’t going to kill people because it crashes all the time. I’m glad there are still systems like System z9 mainframes, System p running AIX, Itanium2 Superdomes running HP-UX and so on. These are systems that are tried and true. And, no, they are not open source. These systems belong—period. What does this have to do with Oracle?
One Port
OK, if you are still with me, picture this. Oracle stops porting to all instruction sets except x86_64—and only Linux. That reduces the cost of producing the database product (by a very small amount I know) and makes analysts happy. It looks more like what Microsoft does. It looks more like what the open source databases do. It looks young and fresh. By the way, I know you can run MySQL on SPARC. Like I say, “sometimes man bites dog.” I digress.
How Would Oracle Pull This Off?
The same way Apple pulled the PowerPC to Intel switch—Rosetta. Rosetta works, we all know that. What not a lot of people know is that Rosetta is Transitive. I just found that out myself. Transitive works. IBM is already using Transitive to woo customers to run their x86 Linux applications on PowerPC Linux. It all starts to make your head swim.
Introducing the High-End Oracle Server of the Future
OK, so there is only one Oracle distribution—x86_64 Linux. That’s it. Well, the way it could end up is that if you want to run the single Oracle port in maximum performance mode, you run x86_64 hardware. How bad is that? Remember, this is the era of commodity processors delivering more than 50,000 TpmC. And Moore’s Law is on your side. Although the current application of Transitive is mostly to bring non-Intel executables to Intel platforms, it certainly can go both ways. How would the likes of IBM, HP and Sun deliver value in their high end systems? It could wind up that the competitive edge between these high end vendors boils down to nothing more than which platform performs something like Transitive better. If you want the really cool things that high end System p offers? Buy it. Load Oracle over Transitive and away you go. Like the power savings of Sun CoolThreads? Buy it. Want to run Oracle on it? You know the drill—Transitive.
I’ll leave you with a final thought. This sort of thing would make more business sense than technical sense. Which do you think has more weight?
Kevin Closson's Blog: Platforms, Databases and Storage 
You say Solaris/AIX/HP-UX are dead. Might be true for AIX and HP-UX, but definitely not true for Solaris. Let me give you two examples,
1. Just put following keyword searches in google and see number of results you get:
95,600,000 for Sun+Solaris
32,100,000 for Redhat+Linux
14,300,000 for HP+HP-UX
20,300,000 for IBM+AIX
With release of Solaris 10 and features like DTrace, ZFS, SMF etc. Solaris is indeed most innovative and “opensource” platform.
2. Look at past announcement from Oracle: http://www.oracle.com/corporate/press/2005_nov/solaris10.html
It may come true if supporting Solaris spends more bucks than earning on it. Its a long way to go. Lets see how “Oracle powered Linux” does for next few quarters and then we may be able to predict.
Atul,
If you read much of my blog you’ll see that I don’ think the technology is dead. Legacy Unix is tried and true technology. Unfortunately, the best technology usually doesn’t win. Volume is the name of the game. Legacy Unix is a dying business. It is unfortunate but true.
It looks more like what the open source databases do. It looks young and fresh. By the way, I know you can run MySQL on SPARC. Like I say, “sometimes man bites dog.” I digress.
Most open source databases run on several operating systems and CPU architectures.
Interesting perspective, Kevin. Wouldn’t surprise me one bit, I’ve seen them take the most moronic decisions over the last 6 or 7 years!
The most distinctive facet of Oracle has always been precisely that it can run on more platforms than anything else.
It appears that Oracle would be giving up on their major competitive advantage – if not the only one – if they ditched all other ports.
Let’s put it this way: if it ever happens, the argument to buy Oracle against anything else becomes pretty thin indeed. If it isn’t already.
Yes, business sense (or nonsense) has more weight. Did you notice how MS is now limiting VMs running under certain OS’s?
Licensing and other business fads (such as vendors underwriting their ports) will distort your one-port prediction.
Gary,
Vendors already pay dearly for ports. I really believe the time is nigh for this. They’ve always wanted to break the shackels of multi-port support. Like I said, HW vendors that wont to prove they are best of breed for Oracle can just get really, really good at Transitive…
But what do I know? 🙂
John,
Yes I know that there are some open source databases running on legacy Unix platforms. Thanks for pointing that out.
Noons,
I honestly don’t think Oracle gets any market share just because of platform depth at this point in the game. Their competition is SQL Server and those folks couldn’t care less if Oracle gets all the Legacy Unix RDBMs deployments. Supposedly people by Oracle for features and ease of management. Having only this imaginary reference port I envision would not change any of that. In fact, it should make the product more stable–no port level bugs!
Like I say in the original post, Oracle’s platform depth through the 80s and 90s is why they are king of the roost today. But I don’t think that fact impresses wall street much any more. For that reason, something will be done.
I don’t agree with you. Linux is not strong enough for High Availavility, and scalability. Linux has born and grown for Intel platform, but Intel IS NOT A SERVER manufacturer, neither a HIGH-END system producer.
There is a big world with thousands of high end systems running on solaris, Aix, HP-UX.. NOT LINUX. Linux is for entry-level, development, small and mid business.. but is useless for the big leagues.
And there are thousands of big customers running Oracle on those platforms.. they won’t switch to linux never ever. Imagine a one-million-transactions-per-hour-system running on a linux box… impossible. You have to buy a one million dollars’ Sun 25k and still pray for it being responding in a decent way.
In the other hand, Oracle core is developed under solaris.
Gaston,
You need to read more of my blog. I’m not drunk on the Linux Kool-aid by any stretch of the imagination.
Dropping SPARC isn’t the same as dropping Solaris.That’s the same as Apple keeping MAc OS.If Oracle’s really thinking of Linux-only solution, I would think of moving to Solaris +anotherDB. Not every application need DB, but they all do need a solid OS.
Kevin,
It [probably] makes more “business” sense to the “Oracle Corp.” than to the zillions of shops running different flavors of systems out there because for them, it would mean spending cost to migrate their running apps to Linux and most of them would not be willing to do that just because Oracle thinks it is better.
Also, w.r.t Linux, most of the shops would have the question “is Linux stable enough to handle high-transaction apps?”. It would also mean that those shops that are running their systems on “big irons” would have to run them on RAC after moving to Linux and that would mean license cost for RAC and Oracle is notoriously expensive when it comes to licensing products like RAC, ASO/ANO, etc.
Amir,
This post is not supposing an imaginary world where you can’t run Oracle on “big iron.” The point is that the bits would not be natively ported bits. If Linux stability is your concern then that is a non-issue because you’d still have Sol,HP-UX,AIX running the hardware.
The whole thing is just an idea anyway. I’m not saying it will happen. I just wouldn’t be surprised if it did…someday.
A quick comment to this old post given that you’ve just tweeted about it Kevin…
I agree with all the above comments about Intel/AMD not being the same as big iron/UNIX HA-wise… even now, nearly 6 years later. But the gap is closing.
I’d never heard of Transitive before. However now it’s not hard to imagine a point where Oracle actually delivers the database as a virtual appliance – based on a paired-down x86_64 Oracle Linux upon which, after the VM bootstrap/reconfigure process, the database software would be ready to go. Now of course that would be an OVM-style VM (unless something drastic happened, like Oracle buying EMC 😉 ) but it wouldn’t be rocket science to load it onto other x86 hypervisors, or – as I think you suggest Transitive offers – running it in a sort of x86_64 emulation container on the big iron of your choice.
However Oracle would still be holding the licence and support cards…