HP iron still haunted by ghost of Compaq
Solaris does ProLiant. Not HP-UX
Comment Hewlett-Packard might have been the acquirer when it merged with Compaq back in May 2002, but when it comes to server operating systems, the Compaq inclusionary philosophy usually prevails even if it does take some time to be turned into action. Like something akin to a decade in the case of Solaris.
Yesterday's announcement that HP would become an OEM partner for Sun Microsystems' Solaris Unix variant is really a Compaq announcement, not an HP announcement. The evidence: Solaris is only available on the ProLiant and related BladeSystem server platforms, which were Compaq's flagship server products and which constitute the bulk of revenues and shipments in HP's server business since HP bought Compaq.
The ProLiant rack, tower, and blade servers are all based on x64 chips from either Intel or Advanced Micro Devices. Sun came to its senses with the launch of Solaris 10 back in November 2004 and decided to not only support x86 and x64 servers from its rivals, but to embrace the idea. Sun mothballed its x86 variant with Solaris 9 during the dot-com bust - a reflexive and stupid decision that let the Linux genie into the Sun data centers of the world.
The earlier Solaris 8 did run on x86 machines and had built up a following, particularly among Compaq shops since Compaq was the volume leader in x86 servers. And a Solaris 9 variant that was open source and available on x86 and then x64 gear might have helped Sun preserve some accounts even if it had to take revenue hits during the last IT downturn.
HP's Integrity server line - the kickers to HP's PA-RISC Unix servers and the machines that are now based on Itanium processors from Intel - were not part of the Solaris love-in yesterday. It's something Mark Potter - who holds the unwieldy title of senior vice president and general manager of Enterprise Storage and Server infrastructure software and blades at HP - was at pains to point out a few times to people listening to the webcast.
And that is a very un-Compaq thing to do, considering that the Integrity machines run HP-UX (HP's own Unix variant and the main alternative to Sun's Solaris and IBM's AIX), Windows, Linux, OpenVMS (the former DEC proprietary operating system), and NonStop (the former Tandem fault tolerant clustering environment). Compaq ate Tandem in 1996 and then DEC in 1998 in an effort to go enterprise with its servers, but in the end, everything ended up at HP. Except the idea that the flagship HP platform should support the same operating systems that the volume systems do. One set of systems should feed into the other, in theory.
Solaris can certainly run on the Itanium processors. A decade ago, as Intel was crawling like a wounded World War I veteran in Verdun battlefield to the initial "Merced" Itanium launch, Sun booted Solaris on these Itanium chips, and three years later, as the bottom dropped out of the IT market and Sun was trying to circle the wagons around Solaris on Sparc, the company was blaming the death of Solaris on Itanium on Intel.
With HP and Intel, the main proponents of Itanium at this point, both being partners with Sun with regards to Solaris, the revival of Sparc on Solaris would seem to be long overdue. Microsoft has relegated Windows Server 2008 on Itanium to a role as a SQL Server engine. Meanwhile, Intel, Sun, Fujitsu, and now HP aren't interested in getting Solaris on Itanium Couple all that with the impending "Nehalem" Xeon EP server announcements and the high-end Xeon EX server announcements later this year, and Itanium is only relevant to a relatively small number of suppliers and customers.
Intel is not exactly out there defending the Itanium's honor as journos, analysts, and companies bash it. Unisys didn't exactly give Itanium a ringing endorsement last week. The job of defending Itanium gets relegated to the Itanium Solutions Alliance, a consortium of Itanium server makers, operating system providers, and application software makers. (I'll be talking to them about Itanium as soon as I can get some executives on the horn).
Potter said yesterday that HP does not have any plans to port HP-UX to its x64-based ProLiant servers. It's something HP has said year after year as the Itanium chip has been delayed and has been pushed into a niche market at the high-end of the server space. Given the state of the economy and the desire to cut costs inside data centers and at Intel and HP, having an HP-UX port to Nehalem chips might come in handy about right now.
The Compaq buy was a smart move for HP, but not porting HP-UX to ProLiant machines might turn out to be a very foolish tactic. This is exactly what Sun tried to do to protect its Sparc server sales during the Solaris 9 generation, and we all see how well that worked out.
Putting Solaris on Itanium is almost certainly a waste of time and money, unless it is relatively easy to compile the code and HP thinks it can make some money chasing Sparc shops with Integrity boxes. Getting HP to officially and globally support Solaris on ProLiant machines is probably a good thing for HP, and it might even be good, in a nominal sense, for Sun.
For one thing, Solaris 10 has been certified on lots of HP ProLiant iron for years. The OEM agreement between HP and Sun allows IT shops that have already standardized on ProLiant machines for certain workloads (very likely for Windows and maybe Linux as well) to get the much-desired "one-throat-to-choke" when things go wrong.
While HP and Sun did not provide any financial details of their OEM arrangement, HP gets to sell support contracts for Solaris and presumably Sun gets a piece of the action. This might be better than the alternative for Sun - which at a lot of shops is facing migrations from Sparc/Solaris to x64/Linux - inasmuch as Sun gets to preserve some Solaris support revenue. But given what will probably be the relatively low volume of Solaris licenses and support contracts HP and its channel will sell, the take will be small for Sun.
With this deal with HP, it may get harder for Sun to argue for its own x64 servers, and you might even say that similar OEM deals with IBM, Dell, and Fujitsu for their x64 iron have hurt Sun's hardware efforts as much as they have helped Sun's software efforts. To its credit, Sun wants Solaris to live in a world where Windows and Linux are the defaults for a lot of new applications and other platforms live by virtue of their installed bases and the legacy applications they run.
That's not a slam, just an observation, and I certainly don't advocate Windows or Linux as the only platforms on which to deploy new applications. Sun has to get every major server maker to support Solaris, even if they are competitors. Even if it hurts Sparc and x64 server sales at Sun itself. That's not the Sun of Ed Zander, but it certainly is the Sun of Jonathan Schwartz.
As part of the OEM agreement, HP is going to assign some software engineers to the OpenSolaris development project, which is how future Solaris versions are being cooked up, and this is also good for Sun. With chip makers Intel and AMD and server makers HP, Dell, IBM, and Fujitsu kicking in, Solaris will be updated more or less concurrently with x64 hardware platforms, and Sun's Software group doesn't have to try to do all this work itself.
One elephant in the room that no one talked about yesterday on the HP-Sun call was the formerly independent Electronic Data Systems, now accounting for about half of the revenues at HP Services and heaven only knows how much server sales each quarter. Sun and EDS have historically been very tight partners, and in the 1990s and during the dot-com boom, EDS installed and managed lots of Sun iron. (Ditto for IBM mainframes, too).
But now that HP controls EDS, it has to be very difficult among EDS accounts to make a case for Sun's x64 iron, and in many cases, it was probably hard to make a case to keep Solaris applications on Sparc iron at all unless big discounts were available from Sun. With Solaris formally embraced on ProLiant by HP and therefore EDS, there's a downgrade path - if you consider moving from Sparc to x64 iron such - that doesn't require a move to Linux even if some companies might want to move x64 iron. This is better than it could be for Sun too. Sun could have gotten nadda.
The real question - and one that so far no one has answered - is what actual impact these four OEM deals have made on Solaris 10 support revenues and license shipments. Is the trend up, down, or sideways? What share of Solaris revenues do they contribute? The public relations of these OEM deals is always good for Sun and its Solaris server partners. But please show us the money. ®