Microsoft isn't the only company affected by Windows' lack of success in the datacenter. Until this year, Windows was pretty much Intel's only hope of getting into the highly coveted market. But the Wintel twosome was unable to reproduce its desktop dynasty in a market where scalability and high availability via 64-bit multiprocessor systems has been the name of the game.

To date, the top end for most Windows SMP installations was either a four- or eight-CPU Xeon (32-bit, Pentium III-class) system -- a toy compared to some of the midrange to mainframe 64-bit SMP systems that IBM, Sun, Hewlett-Packard, and Compaq (by virtue of its Digital acquisition) have in service. As a side note, Microsoft's poster child of multiprocessing on Windows is a 32-way Xeon-based ES7000 from Unisys, based on that company's Cellular MultiProcessing technology (CMP). Despite the fact that it has barely made a dent in the datacenter, any mention of Windows' multiprocessing limitations to someone in the Wintel universe usually draws mention of that Unisys system and its benchmarks.

But during the summer of 2001, several announcements and introductions intersected the Wintel universe. Intel introduced its second 64-bit processor (Itanium) at almost the same time that it announced that Alpha -- its first 64-bit processor by way of acquisition -- would be discontinued later this decade. Microsoft, which once had Alpha on its Windows roadmap, subsequently introduced the first 64-bit version of Windows Advanced Server for Intel's 64-bit architecture (IA-64).

Without question, IA-64 was the ride that Microsoft needed to make it onto the raised floor of today's datacenters, and into the hearts of the managers who run them. The problem for Microsoft is that when IA-64 finally gets to the datacenter, it probably won't be running Windows.

That honor will very likely go to Hewlett-Packard's version of Unix -- HP-UX. According to Ram Appalaraju, HP's director of Unix operating system marketing, HP-UX and IA-64 have enjoyed a long courtship, dating back to the days when HP and Intel partnered to build EPIC (Explicitly Parallel Instruction Set Computing), the foundation for IA-64. Both Intel and HP claim that EPIC picks up where 20th century 64-bit RISC architectures such as HP's PA-RISC, Intel's Alpha, Sun's SPARC, and IBM's Power4 leave off.

Evidence of that courtship was delivered when HP announced the availability of its HP-UX 11i-based, 16-way Itanium systems -- the most of any OS supporting IA-64 -- at the same time that Itanium itself was introduced. Much like their PA-RISC-based cousins that go up to 64 processors, the 16-way Itanium boxes can be configured as either one big computer (single partition), or as multiple partitions with a minimum granularity of one CPU for software isolation and four CPUs for hardware isolation. HP has plans for that scalability to reach new heights next year when Intel ships the next member of the IA-64 family, code-named McKinley. Appalaraju expects to see HP-UX scaling up to 128 processors, when combined with Intel's 870 chipset. According to Intel spokesperson Bill Kircos, McKinley will ship with the 128-processor capability sometime in mid-2002.

Aside from the fact that HP was a partner in developing EPIC, another reason HP was quick to market with a solution is that the company is already on its second generation of 64-bit operating systems, Appalaraju notes. "HP-UX is in its 18th year," Appalaraju says. "When HP came out with HP-UX, the only scalable, highly available, and manageable alternatives were DEC's VMS and IBM's mainframes. We decided back then to make those features a core part of our offering."

HP's investment in a technology it calls ICOD (instant capacity on demand) could pay off now that there's renewed interest in 5-9's (99.999%) availability. According to Appalaraju, ICOD dynamically deallocates bad resources and reallocates the processes that depend on them to especially configured spare hardware. Appalaraju says "it all happens dynamically without bringing the machine down."

HP-UX has a little something for everybody, including shops that have deployed Linux applications. For folks looking to take advantage of some of HP-UX's high availability options without sacrificing Linux, HP-UX comes with a layer of Linux APIs that eliminates the need to remove any Linux dependencies before recompiling an application. Between this capability and HP-UX's partitioning flexibility, HP-UX provides some of the same server consolidation capabilities that IBM does. (Several applications, including ones native to the host OS as well as Linux applications, can run inside of one box, each with dedicated resources.) Like IBM, HP claims that the resulting consolidation can drive down total cost of ownership.

Whereas APIs and recompilation are necessary to support Linux applications now, Appalaraju says that need will go away once HP ships its Linux Application Binary Interface, which will allow Linux binaries to run natively on HP-UX without recompilation. For those who doubt whether HP can pull this off, you should know that HP has a history of building these binary interfaces. Back in the mid-90s, HP built a similar binary interface that allowed NetWare to run on HP-UX. And HP also offers a binary interface for those looking to migrate executables from HP's MPE/ix mainframe operating system to HP-UX.

More importantly, to alleviate any concerns about HP's long-term transition from PA-RISC to EPIC, the IA-64 version of HP-UX includes HP's Aries technology. Aries allows PA-RISC-based HP-UX binaries to run natively on the Intel platform.

After hearing Appalaraju talk about everything HP has done to migrate its existing PA-RISC customers to Itanium, I was left wondering what, if any, bases it didn't cover. Suddenly, I realized why, as I interviewed Intel's Kircos earlier in the year about Linux and Intel, he expressed surprise that I hadn't asked any questions about Unix. Clearly, Intel couldn't wait for Windows or Linux to work their way into datacenters. In the meantime, Unix (and the HP-UX version in particular) looks very much poised to be Intel's first ticket into the datacenter, largely because HP-UX is already there. About the only thing that could hold it back is the uncertainty that surrounds HP's proposed merger with Compaq. But I'm guessing that HP-UX is a shoe-in to survive that merger. Any uncertainty concerning HP-UX is based more on propaganda from HP's competitors than it is on reality.

What do you think? Share your thoughts with your fellow readers at ZDNet TechUpdate's Talkback, or write directly to david.berlind@cnet.com.

Got a great tip? An industry rumor? Or do you want to submit your own column to ZDNet TechUpdate? Send David your submission, and if we use it, you'll be compensated with some of the cool vendor schwag that arrives in our mailboxes on a daily basis.