It all started with a little blog entry by Sun’s new president and COO, Jonathan Schwartz, as he was pondering HP’s most recent quarterly financial results. In that blog entry, posted on August 16 and titled HP’s Problem? It Ain’t the SAP Install, Mr Schwartz started out by saying that he admired HP’s chairman and CEO, Carly Fiorina, because she demonstrates courage, is willing to buck conventional wisdom, and has a titanium spine.
HP was probably very happy to read that bit. But then the other shoe dropped, as Mr Schwartz said this: To me, HP’s problems spawn from the death of . . . their operating system, HP-UX. The ellipsis is his punctuation, it is not clear why it is there. As for the death of HP-UX: most observers didn’t even know it was sick, or even tired.
Withdrawing support for Solaris on X86 to try to protect its UltraSparc server business was one of the least smart things that Sun has ever tried to do, and it was only able to reverse course on this and double down on its investment in Solaris after many top executives at the company left and the new top brass realized that Solaris, not Sparc, was the real reason many Sun customers love the company.
Now, to put it bluntly, Sun is trying to give the impression, in blogs like this one as well as in advertisements and other company statements, that HP-UX is dead because HP is only supporting that Unix variant on PA-RISC and Itanium processors, while Sun is supporting its Solaris Unix on 32-bit and 64-bit Xeon and Opteron processors as well as on 64-bit Sparc processors.
So on September 28, HP sent Sun a cease and desist letter refuting that it is abandoning its HP-UX customers and giving it two weeks to fix the offensive claims on its website, put out a press release correcting these claims, and agreeing to base future claims on valid, objective, and verifiable data.
Mr Schwartz almost nailed a valid criticism correctly in his blog when he suggested that HP-UX won’t even run on HP’s own industry standard servers. The way he said it was only a glancing blow, however. In HP’s view, industry standard can apparently mean any Xeon, Opteron, or Itanium server. In the old Compaq era, it meant simply Pentium or Xeon server.
What Mr Schwartz should have said is that HP-UX has been ported from the PA-RISC architecture to the Itanium architecture, but that HP has not taken a broader approach and made HP-UX available on the 64-bit Xeon and 64-bit Opteron processors. This is a valid criticism, just as valid as the criticism Sun faced many years ago when it decided to kill its Solaris port for Itanium and mothball its Solaris port for 32-bit X86 processors.
When Opteron didn’t exist and Itanium had more advantages compared to other X86 alternatives (and the term X86 is used very loosely there), it was easy enough to say that HP-UX would be Itanium-only for the long haul. HP and Intel were committed to Itanium, as was IBM, the former Compaq, and the prior incarnation of SCO before Linux vendor Caldera International bought it. (Remember the Project Monterrey mix of SCO Unix and IBM AIX for Itanium?)
While there was plenty of criticism for Itanium at the time, it was tough to go against the market inertia of Intel. And had Itanium not been so late and such a poor performer initially with the Merced chips, we might not be talking about Opteron or Xeon-64 today. As long as Opteron had not attained traction with the tier one vendors, HP was credible in its contention that porting HP-UX to this chip didn’t make sense.
But now, three out of the top four vendors sell AMD’s Opteron servers and two of them (HP and Sun) have made substantial commitments to them. IBM’s commitment to Opteron is tepid at best – relegated to a single two-way server aimed at the HPC market – and Dell has yet to get with the Opteron program.
But now, with Intel delivering 64-bit versions of Xeon chips and the Opterons demonstrating advantages compared to both Xeons and Itaniums, putting HP-UX on 64-bit Xeon or Opteron servers is not the stupidest idea in the world, particularly for entry and midrange servers. If Mr Schwartz wanted to point this out, he almost did.
Whether Intel or HP likes it or not, the market may ultimately decide what it really wants is multiple-core Xeon or Opteron processors in machines that scale from 1 to 32 or 64 sockets, with those processors having the lowest clock frequency possible and with the machines being able to run Unix, Windows, or Linux in 32-bit mode or 64-bit mode, side by side.
It will take years to sort this out, of course. Not making HP-UX and AIX available on such 64-bit Xeon or Opteron processors makes this future less possible. But making Solaris, Linux, and Windows available on these machines also pulls the market in the opposite direction from where HP and Intel want to go.
Getting back to the HP-Sun spat. On October 5, Sun responded to the cease and desist letter from HP, saying that Mr Schwartz does in fact believe that HP’s problems spawn from the death of its operating system. The editorial comments found in his personal blog provide an accurate and good faith account of his opinion of HP-UX. Sun then went into a lengthy explanation as to why it would not cease and desist from saying HP-UX is dead.
While Sun is entitled to its opinion, it is ludicrous and reckless to suggest that HP will abandon HP-UX. There is simply too much money to be made from the HP-UX installed base. However, it is not preposterous to suggest that HP and Intel are in the uncomfortable position of explaining and re-explaining their commitments to Itanium, and HP has to constantly defend running HP-UX on that platform.
Until Itanium is more widely adopted, there will be such longevity questions – just as there are for Sparc and Power platforms, incidentally. With a 64-bit Xeon and Opteron that can run both 32-bit and 64-bit X86 code natively, there is tremendous pressure to move in that direction to simplify programming and server purchasing. Admitting this could be the death knell for Itanium, which is why neither HP nor Intel say it.
Having thousands of applications on Itanium, neither wants to contemplate an HP-UX port to Xeon-64 and Opteron and then getting all of those applications lined up yet again on a new architecture. HP does not want to re-engineer its Integrity line of servers so they can use 64-bit Xeons or Opterons. But what HP and Intel want is not the point. Customers will ultimately decide this issue – not HP and not Intel – and they will vote with the budget dollars they spend and the budget dollars they withhold.
