Sun-HP Unix Merger: Beware of Olive Branches Shaped Like Baseball Bats

Published: June 22, 2006

by Timothy Prickett Morgan

One of the great things about the Unix business over the years is that it has had a wide variety of players. No two companies represent the polar opposites of this business more than do Hewlett-Packard and Sun Microsystems. One is the quiet vendor, creating great technology and slowly (over 34 years) building up a vast commercial server customer base, including proprietary and then Unix machines; the latter has only been in the commercial server racket since 1991 (15 years) and shoots off its mouth incessantly about how great its products are--often with good reason.

So many of the Unix players are dead or dying, and today the Unix server business is dominated by the big three--HP, Sun, and IBM--in terms of revenue. These three account for the vast majority of Unix server sales each year, and while Unix has dropped from about half of total server sales in the late 1990s, the Unix collective still accounts for about 35 percent of all server revenue worldwide. To be fair, the open source BSDs and a struggling SCO Group are doing a reasonably high volume of X86 and X64 servers, but they account for very little money.

Given the contraction in the Unix market, and the storied battles between Unix players in the 1980s and 1990s, someone who didn't know the Unix market very well might think that Sun's perhaps half-hearted but full-throated talk of a merging of Sun's Solaris Unix variant with HP's HP-UX variant would be something that HP would consider. The idea is not just laughable, it is technically difficult and economically unattractive. But, because Sun likes to make trouble--even if chairman Scott McNealy is not the one officially making the trouble--the company once again this week extended its "olive branch" to HP to merge Solaris and HP-UX.

In its celebration of 5 million registered licenses of Solaris 10 in the past 18 months, Sun once again echoed the offer that McNealy made to HP president and chief executive officer, Mark Hurd, back in March. "We're again extending the olive branch to HP and offering to help converge the Solaris/HP-UX roadmaps," said Larry Singer, Sun's senior vice president and strategic insight officer, in a statement where Sun was bragging about Solaris 10 shipments. "HP's customer base is already casting their vote by downloading Solaris 10. Running Solaris 10 on X86/X64 servers is a potent combination for our joint customers, and we're eager to partner with HP to make this combo even easier for customers."

The reason why Sun is bringing up the HP-UX and Solaris Unix merger idea is because it has drilled down its database of information concerning those 5 million registered licenses and has figured out that not only were 3.2 million of those licenses running on X86 and X64 servers (about 64 percent of all Solaris 10 registered downloads), but that about 25 percent of these licenses were on HP X86 and X64 servers. Sun says that this is a clear vote that HP customers want to run Unix on X86 and X64 servers.

Well, maybe it is, and maybe it isn't. For one thing, what Sun has not calculated is how many of the HP machines out there are running Solaris. Solaris 10 has been loaded, if only temporarily (since no one, including Sun, knows if a registered Solaris license is in use after it is put on a machine), on about 800,000 HP machines, and the vast majority of these are bearing the ProLiant label. However, as of June 2005, HP had shipped 10 million ProLiant units. Assuming that about half of those units are in the field, and that HP has shipped more than 2 million units in the past 12 months, the installed base of HP machines should be around 7 million units, give or take. That means Solaris has touched about 11 percent of the ProLiant installed base. (Two of those ProLiants are in my home office, and they are not in production. I have another ten Solaris 10 licenses in the office as well.) It is hard to imagine that 800,000 ProLiant machines are actually running Solaris 10 in production, although it is also hard to guess how many are; it is also safe to say that only a percentage of the machines that are running Solaris 10 have support, particularly if the machines are used for infrastructure workloads. If Linux has taught us anything, it is that you don't have to put aging iron on paid-for tech support even if you do put it to use in production.

By comparison, HP had sold 1 million real and new Linux servers, with actual paid-for licenses, through June 2005, and with Linux accounting for roughly a quarter of server shipments these days, that would imply that HP has probably added another 500,000 Linux machines since last June. These are an approximate 1.5 million real servers, in production, not registered licenses of software that might be in production or not. You don't pay for Linux and then not use it.

While Sun is to be commended for opening up Solaris with the OpenSolaris project, for freely distributing Solaris 10 on two-socket and smaller servers, and for charging a modest fee for tech support for Solaris 10, the fact that so many machines in its 5 million-strong Solaris 10 license database have an HP label makes perfect sense--HP represents about a quarter of worldwide server shipments and the installed base. How could it be otherwise?

The numbers that Sun is shouting about lead me to a few interesting questions, ones that I suspect Sun will never answer. How many of those Solaris 10 licenses are at shops with Sparc machines that are looking for ways of moving their Solaris workloads off Sparc and onto cheaper (perhaps already paid for) X86 and X64 iron? What do you anticipate the effect of the broadening of the "Galaxy" Opteron-based server line and the advent of the "Niagara" T1000 and T2000 servers to be on more general purpose Sparc server sales? When will Sun ship more Solaris licenses on its own Galaxy boxes compared to Linux?

Maybe Sun is serious, as its executives claim it is, and this is not just a PR stunt, as I happen to think it is. And I think Sun's olive branch is not a symbol of friendship. It would be a weapon, if anyone really wanted this Unix merger to happen and Sun could figure out how to wield it.

Merging these two Unixes in such a way that the combination would run applications originally tooled for Solaris on Sparc and Opteron or HP-UX on PA-RISC and Itanium would be very difficult. Every customer would feel lots of pain, and it is pretty safe to say that no application would run unmodified without the use of some pretty serious hardware-based emulation or some other magic I have never heard of or seen. And economically, HP has nothing to gain by working with Sun. There is far more profit for HP in serving its existing HP-UX customers, who spend about as much on Unix-related stuff as Sun's customers do. (Somewhere south of $10 billion annually for each vendor.) HP is supporting Solaris 10 on its new BladeSystem blade servers, and will install it on ProLiant tower and rack servers, too. Sun might want to not antagonize HP and be happy that Solaris 10 is not verboten, ignored, or bashed by HP.

When asked about all of this, HP's top brass in its Enterprise Storage and Servers group didn't even really want to dignify Sun's merger proposal with a response, and they were, as usual, polite about it. "We do not want to have a divisive argument with Sun in the press--as much as you guys might like it," explained Don Jenkins, vice president of marketing for HP's Business Critical Systems unit, which sells the Integrity, AlphaServer, PA-RISC, and NonStop server lines.

If Sun wants to be a diplomat, instead of a wiseguy, it could broker a settlement between IBM and SCO over Unix intellectual property and other issues before that famous lawsuit goes to trial next February 2007. It could then convince SCO to let all of the Unix code go open source, and then work with IBM, HP, the BSDs, and SCO to propose a single Unix standard that runs across all server architectures--including the Itanium architecture Sun has mocked so much. If Sun wants Unix to survive, beat Linux, and thrive, this might be a great way to do it. But I wouldn't hold my breath waiting if I were you.

RELATED STORIES

Sun Goes After HP-UX Base, Again and Again

McNealy PR Stunt Number 105: Merge Solaris and HP-UX