Monday, April 4, 2011

A bad case of S&D!

Paraphrasing local bedding shop in the west, “You’re killing me, Larry!” But it’s not so much his posturing as it is Larry’s presumption that he enjoys special relationships with every vendor up and down Silicon Valley!

It’s as if I can never escape infrastructure projects, or escape being embroiled in technology obsolescence discussions. Setting up offices for Margo and me in our Boulder home has been quite an ordeal and this morning we had a longtime audio / video specialist, Brian, return to oversee the pulling of Ethernet cables, some coax and other related components needed to support dual working environments. The picture
 above is of Margo who bravely soldiered on while everything was in chaos for several hours.

Watching new networking infrastructure being installed and watching older equipment being carted away, I was reminded of how when I worked for Tandem Computers I used to enjoy providing the product management perspective on communications and networking to customers and prospects where often the message was to forget about what you had already purchased! It was already legacy! In full flight I could address issues to do with SNA/APPN, Asynchronous Transfer Mode (ATM), and the OSI stack without missing a beat.

However, it was while I was so engaged that one day a good friend of mine from development, Brad, suggested I had become the master of S&D; in his words, “Substitution and Distraction” with a performance akin to that of an illusionist, but with a twist.

Whenever a question came up about Tandem capabilities not yet in plan I would transition to a different product or feature and in so doing, present an opinion so energetically and enthusiastically that I succeeded in distracting the audience completely. Rather than having a flaw exposed, I could be relied upon to substitute a related topic so smoothly that to all present, the original question was long forgotten!

Perhaps this is the perfect segue to the issue that’s arisen with Oracle’s CEO March 23rd, 2011, announcement that “Oracle has decided to stop developing software for Intel’s Itanium chip because it thinks the processor is ‘nearing the end of its life.’ Hewlett-Packard screams and Intel reiterates support for the Itanium” was how ZDNet “Between the Lines” blogger summarized Larry Ellison’s latest proclamation.

HP and Intel had every right to be upset and then it really hit me; this was Larry being Larry and facing some major issues himself he, too, was simply resorting to a little S&D!

Should HP NonStop server customers be alarmed in any way? Could the reliance solely on Intel’s Itanium chip prove to be problematic in the future?

“HP and Intel have a long standing relationship of developing enterprise solutions for the mission critical and high availability markets. Together, we have demonstrated our commitment to our customers and partners, with new chipsets delivering increased performance, lower power consumption and maintaining best in class technology,” according to Randy Meyer, Director of Product Management, NonStop Enterprise Division (NED), HP. “We continue to develop and innovate, exploiting the Itanium processor family and increasing the available options in our NonStop portfolio.”

I happen to be firmly in the camp of there being no immediate or midterm issues for HP NonStop server customers and prospects. I have sat through the product roadmap presentations enough times to understand the specific chips NED will use and the timeframes involved.

However, and possibly more relevant for HP NonStop server users, I am of the opinion that Larry, the master of S&D, has a lot to hide. And the more I talk to technologists, the more I am convinced that there are at least three possible issues Larry faces within Oracle!

Larry plans to sell the Sparc chip business and needs to generate more hype and demonstrate its future profit potential. Expressed as simply as I can, when it comes to chips of this complexity, where Sparc is trying to catch up to IBM’s Power chip, let alone Itanium, the potential to be “one-of-two” certainly looks a lot more attractive than being “one-of-three” – basic math tells me there’s something to be gained from this outcome.

Why should Larry sell the Sparc chip business? That’s the $5 Billion (maybe more) question and its roots are deeply tied to the cost for the next chip spin of Sparc which hast to happen sometime under his watch. Ever considered the price of a new chip fab factory, plus the associated R&D expense?

For a company like Oracle, doing well admittedly with quarterly net income hovering around $2 Billion, such a “roll of the dice” is a little too rich, even for Larry. So, Sparc just has to go, and it has to be sold with a premium because Larry has to win!

However, Larry also has a Plan B, and this is the second issue; should the thin veneer he has thrown up around Itanium be penetrated and his plans to sell Sparc crumble, then he will bring IBM into the ruckus as well and really go after both the Itanium and Power programs. If he’s to be left with x86 processors as his only future platform, then he has to start the heavy-lifting associated with neutralizing the oppositions chip sets!

After all, Larry is pushing hard to sell the full stack once again, as he envisions Oracle becoming a one-stop shop! It’s hard not to miss how envious Larry is of the old IBM. Perhaps the one comment he made as he took a shot at the Itanium chip that he would like to take back is when he expressed envy over IBM’s mainframe model from decades ago.

Readers of UK publication, The Register, on March 24th, 2011, could hardly have missed the commentary provided by Dan Olds of Gabriel Consulting, when he suggested that “to me, this is the reality of what Ellison meant when he said that Oracle wants to be the IBM of the 1960s. Oracle wants to have the incredible margins that IBM enjoyed back then. It wants to have that lock-in that IBM had in the days when there were few alternatives and even fewer standards that would allow customers to easily move from vendor to vendor.”

Expressed a little differently, and with an admonishment, writer Jonathan Eunice posted on April 1, 2011, of how “Oracle CEO Larry Ellison has described the mainframe era and model in glowing terms; if he's hell-bent on going there, and on pulling customers into Oracle's ‘full stack’ fold, then Oracle's traditional commitment to heterogeneous multi-vendor support goes by the wayside. If I were an Oracle customer running on any non-Oracle hardware, I'd be asking for long-term support and software update assurances, in writing, with clear and enforceable penalties for non-performance. Verbal reassurances would no longer be sufficient.”

Left with just an x86 chip then yes, to drive home the “full stack” where he doesn’t hold all the cards is not the typically “winning-hand” that Larry would want to hold.

Unfortunately, that’s not all that he’s facing; there’s more! There’s a very serious concern, and the third issue that I see – Oracle, as a data base, is flawed and it’s becoming increasingly obvious to large users.

At issue is the strong tie Oracle has to Symmetric Multi-Processing (SMP) architectures. It can be argued that it has been the exploitation of SMP that has helped Oracle to achieve the success it has enjoyed to date, but when data bases grow too big, and clustering is pursued, SMP really gets in the way! Shared nothing architectures, with the relationship they enjoy between processors, the disk processor (DP2) and the disks, support linear scalability whereas clustering SMP nodes usually peaks at around 3, perhaps 4 – certainly well short of the 100 you find in the text books.

Of late, several very large customers of Oracle have moved to NonStop and in the weeks leading up to HP Discover event I am sure we will hear a lot more about them. But I sense this is the start of a gradual shift in emphasis with NonStop as the answer to a secure data base that can simply scale practically without limit, and without Larry, will become extremely newsworthy! Perhaps these flaws in Oracle are really what have Larry anxiously practicing S&D!

In the meantime, as HP’s Randy Meyer went on to add, “NonStop customers continue to rely on HP’s NonStop Blade technology. The value proposition has been proven time and time again. Leveraging the RAS features that are fundamental to Intel’s Itanium chipset, NonStop continues to deliver world class availability and scale for customer applications that demand true 24x7 availability.”

As for my own home office data center, the installations have been completed and the systems are up and running and I am free to continue writing. I will keep watching the headlines though, as I have to believe we haven’t heard the last from Larry on the topic of Itanium. However, the shock value has now diminished and Itanium users are no longer uncertain about the future so I have to wonder, is it now IBM’s turn to appear in the cross-hairs of Larry’s sights?


MartyTInOZ said...

No matter what HP or Oracle say about Itanium, what matters is what Intel says. If you example the 2011 Intel public processor road-map (link provided below) you'll note that Itanium doesn't seem to have much of a described "future" beyond Poulson. That may or may not be the case internally within Intel, but Itanium customers/users won't know that until Intel does decide. (search for "Itanium")

Richard Buckle said...

Poulson! Yes, that's definite as is Kittson on the roadmaps I have seen of late andin my own talks with Intel then I can see a clear picture right up to the next decade ...

But for sure, things can change but I'm pretty happy with what I am seeing in terms of exchanges between HP, NED and Intel. It's cool ... as indeed, Kittson is!

But thanks Marty - this needs to be all aired within the community and every side heard; and if this is what happens then we all have Larry to thank, right?

Thomas Burg said...

I wonder what the real options for someone running Oracle on HP UX Integrity are:

(1) move to Sun Sparc (making Larry happy)

(2) move to Sybase on HP UX Integrity(making Leo happy)

(3) move to Oracle on Linux on HP x86 hardware (making both Larry and Leo happy?)

(4) Move to SQL/MQ on NonStop (only half kidding)

(5) Pray that Larry gets to his senses

Neither of these sound attractive as everything except (5) implies a hefty reinvestment in either HW or SW - did I miss any feasible options ?

Anonymous said...

You may have missd a thing: Oracle's (expensive) Exadata machines. Since most Itanium-based dsatacenters use OS's that can support Oracle software (databases, product suites) Oracel won''t sell this hardware: "Either buy Exadata of downgrade to X86".