Engineered systems, engineered financials
The big rationale for Oracle creating engineered systems based on Sun hardware running Solaris or Linux, tuned to run best on its own stack, was that this makes the job of supporting its software easier and cheaper. This is all good, and it will help to make Oracle a more profitable company if it succeeds. But supporting Oracle's diverse software base on myriad operating system and middleware platforms is expensive. And someone has to pay for this best of breed custom stacking that customers like to mix and match things across vendors for technical and political reasons. We happen to think it will be Oracle's partners that will end up footing the bill.
So all of Oracle's talk about Itanium being a dead-end product may not be the point. Getting Apotheker's attention - and enough that he will get out that big old corporate HP checkbook out of his pocket - is probably what this is all about. And Paul Otellini, the CEO at Intel that wants to see Itanium systems sell for at least the next five years (and probably more), may have to get out his checkbook too. If companies want Oracle database, middleware, and application software to run on non-Oracle operating systems and hardware, it looks like they are going to have to pay. Logically, this will apply to IBM's AIX as well as to Red Hat's Enterprise Linux and Microsoft's Windows, if this is indeed the Oracle strategy.
In the transition from Oracle 10g Release 2 to Oracle 11g, Oracle already started trimming the platform support it offers. Take a look for yourself here.
Oracle 11g runs on Windows (32-bit and 64-bit); Linux on x86 (32-bit) and x64 (64-bit); Solaris on x64 and Sparc (both 64-bit), HP-UX on Itanium and PA-RISC (both 64-bit as well), and AIX on Power (64-bit). That's it. Here's are the platforms that supported the Oracle 10g database but didn't make the 11gh cut: Linux on Itanium and Power, Windows on Itanium, Mac OS on x64, HP Tru64 Unix for PA-RISC, OpenVMS for Itanium and Alpha, Solaris on x86 (32-bit), z/OS and OS/390 for IBM mainframes as well as the BS20000/OSD for Fujitsu mainframes all have had their support pulled. The 64-bit version of Linux for IBM mainframes is available, contrary to what this story originally said; the download is available for Oracle 11g R2.
None of these pulled platforms were big money makers for Oracle, but the question you have to ask yourself is this: did the lack of Oracle 11g support help kill Windows and RHEL on Itanium? We think it did.
So, if HP wants HP-UX to live, it is going to have to work out some kind of deal with Oracle to make sure its code runs on the future Itanium chips. Logically, what applies to HP-UX applies equally well to AIX, RHEL, and Windows over the long haul. And just because Oracle and its frenemies don't make announcements of cooperation on development, testing, and support, that won't mean such deals haven't taken place or won't in the future.
And equally importantly, Intel and HP have to come clean on what the long-term roadmap is for the Itanium chip family. Outting the feeds and speeds of next year's eight-core "Poulson" chips and throwing out the "Kittson" name is no longer sufficient. If there is a roadmap for Itanium, or as plan to converge the Xeon and Itanium processors that will allow HP-UX and its applications to run unmodified and without a huge performance penalty, Intel has to start talking. Like yesterday.
The alternative - and a much more interesting one as far as I am concerned - is for HP to say to hell with Oracle, to buy Red Hat for its operating system and EnterpriseDB for its Oracle-compatible implementation of PostgreSQL. And kick Oracle right back. ®