Datacenter short list getting shorter

2001 could be remembered as the year that datacenters everywhere went back to the drawing board, judging by the e-mail response to my various mentions of Carly Fiorina's plans for merging HP and Compaq.

While much of the analysis has focused on what the proposed merger means for this Journada and that iPaq (I prefer the new Journada), or this Vectra and that Deskpro, most IT decision makers appear ambivalent when it comes to these rather "disposable" solutions. I may like the Journada better, but the truth is that I could live with an iPaq (and do). And a year from now it won't matter because Moore's Law guarantees me the delivery of an even more capable PocketPC device than either of the two. These are not earth-shattering issues.

However, the HP-Compaq merger is turning the world upside down in datacenters, where decisions often cost millions of dollars and managers, who typically subscribe to the "if it ain't broke, don't fix it" theory, don't like to see much of anything change.

A decade ago, there was plenty of big iron to go around. A tempting buffet of scalable and reliable pairs of hardware and operating system software whetted the appetites and budgets of CIOs. But industry consolidation has altered the landscape substantially. If the proposed HP-Compaq merger goes through (a big "if" now that the Hewlett family is opposing the deal) and HP CEO Carly Fiorina sticks to her plans of moving all assets to the 64-bit Intel architecture, the remaining hardware menu will be narrowed down to three choices by the end of the decade: Intel's IA-64, IBM's Power4, and Sun's SPARC. Not a lot to choose from.

Traditionally, selecting a hardware platform was secondary to choosing an operating system. But the sudden narrowing of hardware choices is sending many datacenters back to the drawing board. The HP-Compaq merger is a particular source of consternation for some datacenter managers. At Gartner Symposium, Fiorina made it clear that not everything would survive the merger.

Although she was merely stating the obvious, Fiorina's candor reaffirmed the significant uncertainty over the future of the merged company's industrial strength operating systems (OpenVMS, MPE/ix, HP-UX, and Tru64, and NSK). The news came at a time when some of Compaq's Alpha-based OpenVMS and Tru64 customers were already reeling from the Intel's June 28 announcement to kill Alpha. Most of HP's PA RISC-based MPE/ix and HP-UX customers were less affected by the forthcoming hardware switch. They saw the writing on the wall in the early 90s when HP partnered with Intel to co-develop EPIC (Explicitly Parallel Instruction-Set Computing), which now serves as IA-64's foundation.

But, in datacenters where the big money is spent, Alpha and the two operating systems that ran on it (Tru64 and OpenVMS) were widely revered for their elegance, scalability, fault-tolerance, and overall dependability. That confidence was rocked by news that the battle-tested 64-bit Alpha would be replaced with its Intel IA-64 counterpart. The IA-64, which has just started shipping, has yet to demonstrate any real-world scalability or the ability to survive the rigors of a datacenter. While OS executives like Compaq's Mark Gorham (see VMS story) and HP's Ram Appalaraju (see HP-UX story) are optimistic about the transition to IA-64, my correspondences with Tech Update readers indicate that not all HP and Compaq customers share that enthusiasm.

Leaving two proven and functioning 64-bit hardware platforms for a younger, less-proven entry was enough to raise an IT manager's eyebrow. Add to that the uncertainty over which operating systems will survive the merger, and now the other eyebrow goes up. Managers are facing a double whammy: unknown survivors running on an unproven platform. The uncertainty comes at a particularly interesting time because the Web services trend could renew interest in both scalability and reliability.

Of course, just when technology decision makers are sizing up their industrial strength computing options, the masters of FUD (fear, uncertainty and doubt) are jumping all over the HP-Compaq merger. Whereas mergers often represent the loss of one competitor to a market, Sun CEO Scott McNealy joked about how this merger represented the loss of two competitors. Michael Dell (not exactly a household name in the datacenter) swooped in for his share of the spotlight too.

Hype and FUD aside, if I had to guess, the long-term, datacenter-proven survivor of the merger will be HP-UX. Shortly after the merger is complete (if it completes), HP will likely announce that OpenVMS and NSK will survive as well. But I have my doubts. If all goes as planned, HP-UX, with its heritage of scalability and reliability, could get a big boost from IA-64. Those plans, of course, depend on Intel delivering a datacenter-ready platform. While I normally don't believe things until I see them, I'll go out on a limb and predict that Intel will deliver. (I can already hear Intel spokesperson Bill Kircos saying, "That's no limb, David!") The EPIC platform has been in development for almost a decade and IA-64's future was almost certainly secured after Compaq sold Alpha and the team that built it to Intel.

The operating system already supports Itanium in a 16-processor configuration (the most of any OS) that can be set-up as one or multiple partitions. HP plans to support 128-processor configurations as early as next year when Intel ships Itanium's successor (code-name McKinley). To address the failover capabilities of NSK, HP-UX is outfitted with ICOD (Instant Capacity on Demand), a technology that can dynamically re-allocate hardware such as a spare processor (regardless of whether it is in the same or different chassis) if something in production fails.

HP-UX already runs MPE/ix binaries natively, providing the MPE faithful (thanks to all of you for writing) with a smooth migration path to something just as scaleable and reliable--if not more so. HP-UX is unlikely to disappoint Tru64 users, whose scientific and technical applications thrived on Alpha's performance.

Finally, there's OpenVMS. After first writing its obituary, I was so encouraged by an interview with Compaq's Mark Gorham that I decided to give OpenVMS a chance. Tech Update's OpenVMS users practically rejoiced. However, OpenVMS, the pure oxygen of operating systems, has another major problem mentioned by several readers: finding people to run it. Proponents tell me that it's one of the easiest operating systems in the world to run. I've heard stories about systems that haven't been re-booted in years. But the truth is that OpenVMS is a mid-range to mainframe-class operating system, and few CIOs will run it without at least one guru on board. Those gurus are few and far between, and they're growing fewer and farther. The new crop of college grads is racing to enter the ranks with backgrounds in other, higher-profile operating systems. As a result, I'm not de-listing OpenVMS, but my newfound optimism is downgraded to "very guarded."

Still, this is just my crystal ball. As with any fortuneteller, your mileage may vary. It can't hurt to check out your alternatives, especially now that the Web services trend is grounds for reassessing the bottom of the application stack (the host and its OS). Several scaleable and compelling offerings based on the remaining shakeout survivors -- IBM's Power4 and Sun's SPARC -- were recently announced, more are coming, and both companies offer something that datacenter managers appreciate in an OS and processor roadmap: stability.

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.