back to article Oracle to HP: 'Liar, liar, pants on fire'

The tit-for-tat between Oracle and Itanium chip partners Intel and Hewlett-Packard rolls on. Oracle has released a statement saying that contrary to claims from Intel and HP, its decision to halt Itanium software development is not putting enterprises and governments at risk. "Just the opposite is true," Oracle said with its …

COMMENTS

This topic is closed for new posts.
  1. Anonymous Coward
    WTF?

    LarryWorld

    Dealing with Oracle is becoming like negotiating an arms deal with the USSR. "What's mine is mine, what's yours is negotiable." I wonder what Oracle's plan B is?

    1. Allison Park
      Boffin

      Mark Hurd was well aware of the end of Itanium and now Oracle knows

      This is not about vanity or revenge.

      This is about HP firing the person who knows the true future of Itanium and that person is now the president of Oracle.

      Oracle has no plans to support the next chip which will have ZERO per core performance increase and does not have a future. There is no reason to support Poulson when Kittson is nothing more than a minor speed bump.

      I hear HP is not even going to have a follow on to the SX3000 chipset so all the QPI speed increases for Poulson wont even show up in the Superdome2.

      Unfortunately HP customers pay for Mark Hurds R&D cuts...and now have to pay for Mark Hurds Oracle software cuts because of his HP hardware cuts.

      This also shows you that Oracle has no intention of doing the same to IBM's Power chip as it is growing, has a long future and is the real only alternative to Nehalem....well unless the T chip can survive long enough to run databasts

      1. Matt Bryant Silver badge
        Stop

        RE: Mark Hurd was well aware of the end of Itanium and now Oracle knows

        "....he next chip which will have ZERO per core performance increase...." Strange, so you've benchmarked a chip that isn't even out yet? Do you have a timetravel machine? 'Cos if not, you're just talking out of your rectum. Poulsen will be a die shrink to 32nm, so just that alone hints at better performance. Then there's the doubling of the number of cores per CPU, which will mean twice as many cores per socket in the same servers customers buy today. Then, unlike IBM (who just concentrate on cranking up the core clock without sorting the bottlenecks in the rest of the chip), Intel will have added even larger cache (54MB, waaaaay more than Pee7, and probably more than Pee8) and mutlithreading and parallelisation tweaks. All of which points to a good performance boost. Maybe you should try reading up on what Intel plans before your next post.

        "....I hear HP is not even going to have a follow on to the SX3000 chipset...." Really, from whom? Not Intel or hp, I wager. Don't tell me, you heard a voice telling you this whilst you were riding on your timemachine, no?

        "....This also shows you that Oracle has no intention of doing the same to IBM's Power chip...." And what exactly is to stop Larry doing the same to Power? After all, he sells more Oracle licenses on Integrity than he does on Power, so surely it would hurt him all the less to do the same to Power.

        1. RegGuy
          FAIL

          And what exactly is to stop Larry doing the same to Power?

          "....This also shows you that Oracle has no intention of doing the same to IBM's Power chip...." And what exactly is to stop Larry doing the same to Power? After all, he sells more Oracle licenses on Integrity than he does on Power, so surely it would hurt him all the less to do the same to Power.

          Doh: DB2.

          Oracle already frightened many customers into IBM's hands at their buyout of Sun. If they don't commit to Oracle on Power customers will simply migrate to DB2. Why would they change their Power platform because Oracle is no longer supported on it? There's a lower business risk just changing one component, and if that means everything is IBM then it's single point of contact, to boot (one supplier to kick).

          Doh.

  2. K. Adams
    Black Helicopters

    "... [Intel's] plans to replace Itanium with X86 are already in place."

    It's a self-fulfilling prophecy:

    1. Larry Ellison and Oracle believe Intel and HP are going to stop development on Itanium and get out of that market.

    2. Larry Ellison and Oracle drop Itanium as a supported platform.

    3. Having lost 25% to 40% of its Itanium customer base, further development of Itanium by Intel and HP is no longer profitable.

    4. Intel and HP stop development on Itanium and get out of that market.

    5. Larry Ellison looks like a genius prognosticator, and Oracle profits handsomely by finding a home for dislocated Itanium-hosted, Oracle-based apps on Oracle hardware.

    So Intel and HP will end up being "liars" in the end, but not by choice.

    Reminds me of that game Martin (Robert Redford) and Cosmo (Ben Kingsley) used to play, as depicted in the movie "Sneakers":

    -- -- Cosmo: Posit: people think a bank might be financially shaky.

    -- -- Bishop: Consequence: people start to withdraw their money.

    -- -- Cosmo: Result: pretty soon it is financially shaky.

    -- -- Bishop: Conclusion: you can make banks fail.

    -- -- Cosmo: Bzzt. I've already done that. Maybe you've heard about a few? Think bigger.

    -- -- Bishop: Stock market?

    -- -- Cosmo: Yes.

    -- -- Bishop: Currency market?

    -- -- Cosmo: Yes.

    -- -- Bishop: Commodities market?

    -- -- Cosmo: Yes.

    -- -- Bishop: Small countries?

    -- -- Cosmo: With luck, I might even be able to crash the whole damned system...

    And that's a shame, because Itanium, while definitely not a "household name" as far as server processors go, was finally starting to grow some real legs...

    1. This post has been deleted by its author

    2. asdf
      Flame

      almost grasshopper

      Your chain of events may match what Larry has in his head but in reality in step 5 people will go to x86_64 or elsewhere instead of slow as molasses POS garbage architecture that is SPARC. Still Larry at the minimum hopes to kill off a competitors platform that is costing him $ to support.

  3. The Great and Powerful Oz

    Is it just me, or ...

    The article says " Red Hat Enterprise Linux 6 already does not run on Itanium and whatever follows Microsoft Windows Server 2008 R2 to market won't either."

    That's not really relevant.

    Did the author miss the point that Microsoft dropped Windows support of IBM's Power chips years ago, back in the early NT days? I don't think they ever supported SPARC. I believe Red Hat dropped SPARC support in 2000. Those events didn't exactly spell instant doom for either of those architectures, or cause Oracle to drop support for them. Implying that HP's systems don't have software developer support because Microsoft and Red Hat dropped them is a misdirection. Non-support from outside OS vendors, declining market share, etc are much more valid criticisms of SPARC at this point than of Itanium. Look at how many large businesses run SAP on HP-UX, for instance. there's plenty of support out there. This is an attempt by Oracle to shore up falling hardware sales.

    1. Gaius
      Unhappy

      SPARC

      They did support SPARC tentatively, but endianness issues in the HAL meant it could never be made to perform and it died a quiet death. Had it running on an UltraSPARC 1/143 at one point.

  4. Steve Davies 3 Silver badge
    Stop

    Children, Children!

    Go and sit on the naughty step.

    Oracle you really should take a leaf out of IBM's manual. The like you sell Hardware that runs Unix. Do they throw their toys out of the pram when a customer wants to run DB2 on another hardware platform? No they don't.

    This is Oracle's, sorrt Mr L Ellison's Vanity pure & simple.

    1. This post has been deleted by its author

      1. Paul M 1

        Title

        Considering that Sun themselve were pretty ambivalent about Solaris on x86 at one point it seems odd to criticise a major competitor for a slack of enthusiasm.

      2. Gaius
        Linux

        x86

        That doesn't mean anything; it is only very recently that Oracle themselves have started supporting Solaris x86/x64 properly, it used to lag so far behind the SPARC version for feature releases and bugfixes that no-one would ever gamble their business on it.

        I'd go so far as to say Red Hat's entire business is based on exploiting this issue, since they were for a long time the only viable way to run Oracle on commodity hardware.

      3. Anonymous Coward
        Linux

        Huge difference

        less than 1% of customers run Solaris on x86...why would anyone support it. Even Oracle does not have Solaris on Exadata....except for the brochure

        HP-UX is only on Itanium

        Comparing the two markets is ludicrous

  5. deadbeef
    Pirate

    DB2

    Wouldn't surprise me if IBM decided to join the party and yank DB2 from Itanic just to stir things up even more so that they can mop up all the customers from the sinking ship. Probably more profitable in the long run rather than pouring gobs of money into further developing DB2 for a dead platform with such a small market share. Probably what Oracle were thinking.

  6. ShesAllOverIT

    Childish decision

    In an article posted at this site by Dan Olds of Gabriel Consulting, Mr. Gabriel correctly states that "Oracle .... has reinforced its reputation as the North Korea of the server business … only without the charisma and cool Texaco flag."

    Dan Gabriel is spot on in his analysis of the situation, and I encourage all to read the entire article.

    I believe that Larry and Mark will soon rue the day they made this heinous and utterly childish decision.

    1. This post has been deleted by its author

      1. Anonymous Coward
        Grenade

        The thing that guts me...

        ... is that as a customer of "Oracle" ... ie. Sun ... is that all this power play has nothing for the customer. It only makes our lives pathetically painful by having to move from one system to another, simply because Ellison wants to crush HP and wreck a perfectly reasonable chip technology.

        As a customer, this power play, supplier toy-throwing-out-of-pram means that when it comes time to switch server for our major database systems, it darn well won't be Oracle that we stay with. Heck, we're only in to them for the Sun hardware and Solaris ... our databases are Ingres !!!

        We want a supplier who supports us and moves with what we, as a customer, want and need. We're here to ensure that we get systems that provide value for money ... and the moment a supplier looses sigh of this, they're history.

        Ellison ... when you're inherited kit gets wheeled out the door in about a year or twos time, I'll be breathing a sigh of relief.

      2. Steven Jones

        @Oliver Jones

        What on earth makes you think there is any real personality type difference between those who thrive in capitalist and authoritarian regimes? Generally speaking, it requires a degree of ruthlessness and enormous ambition to do either. It's no accident that the hugely rich Russian oligarchs were also the ones that were well represented under old regime. Also the matter of egos can't be missed out - it's no coincidence that rows and break-ups get dirtier the higher up the management chain.

        However, that's not the point here - quite apart from the commercial matters involved here (and this move certainly undermines competition), it's a real slap in the face for a lot of big companies running big-iron HP-UX. It's hugely disruptive and expensive to migrate very large databases on mission critical systems. Oracle will not have pleased many of their big customers with this move.

        1. John 62

          Das ist Kapital? Alle ist Kapital!

          everything is capital

          some people prefer to trade money, or assets that can be valued in terms of money, and some people prefer to trade in less liquid things, like personality, ideas, reputation and plain old force of will and often there's a mix of everything involved.

  7. pan2008

    when oracle can beat other databases...

    When oracle can beat this I will be worried by the deal... Has anyone tried to beat the real database test with the 40 tables instead of the 8 they are falling over themselves to beat from 1992? Lets be Men Mr Ellison, give your best shot and stop hiding in your boat...

    http://www.tpc.org/tpce/

  8. Anonymous Coward
    Anonymous Coward

    Not a surprise

    If Oracle didn't own Sun, none of these conspiracy articles would be popping up on the internet. Everyone but Matt B and his followers see the writing on the wall. Itanic is going away. If Oracle could make money on it, they'd stay on board. Like most companies, they like to make money. If Oracle was doing it for their own server sales, it would be after they released the T4 or next Sun/Fujitsu CPU. Not now. It's purely about making money off their investment of developing and supporting code on Itanic and they found it lacking. Pure and simple.

    1. Jesper Frimann
      Headmaster

      Hey, I am not a Matt B follower !

      But I am a long time *NIX fanboi. And the move by Oracle is bad for us, "The Client's". This will mean less competition and it will also force a lot of clients to do migration (Ok we'll make money on that one), but this messes things up quite a bit. So I think it's a bad move for us clients.

      As for the timing, then this announcement has really no effect right now, but it'll be more of a snowball running down a snow filled hill, it'll gain momentum in the years to come. Hence the timing by Oracle is just right.

      // jesper

  9. John Savard

    Last Major Company

    Does IBM support the Itanium platform with its current releases of DB2, contrary to what Oracle is claiming, or not? If it doesn't, one can hardly fault Oracle.

  10. John Savard

    IBM still loves Itanium

    The current version of DB2 from IBM, version 9.7, still supports the Itanium, at least running HP-UX. However, while version 9.1 also supported the Itanium running Linux, that was dropped by version 9.5 at least.

    If Oracle was the last major company to support the Itanium, then, I guess that IBM isn't a major company. That will come as news to many there.

  11. Anonymous Coward
    WTF?

    What a silly game this turned into...

    Tim, as a respected industry journalist, can you tell me who the biggest selling vendors of servers are? By volume, by value, whatever? Does Proliant come in the top two or three still? How much bigger in volume and value is Proliant than IA64? So whenever (not if) IA64 rides off into the sunset, HP are still well positioned as a server vendor (if, that is, any of their potential customers still trust them - and many would sympathise if they didn't).

    How big is the biggest current Proliant in terms of memory and processors, Tim? How does that compare with high end Integrity?

    Look at the hardware capabilities of x86-64 in high end Proliant and it becomes obvious that folks buying Integrity are, with a few exceptions, buying it *despite* IA64 not because of IA64, They're buying it because of investments in and capabilities of HP-UX, Non-Stop Kernel, and VMS. The exceptions are those folks who want more memory and more processors than you can get in a single Proliant 580series (32 cores and 1TB with Intel, or 48 cores and 512MB with AMD. Not quite in Superdome class yet, but how many people need that much?)

    From recent coverage it seems as though NSK and VMS no longer exist as far as Big Iron HP are concerned, but they do, and from an outsider's point of view a port to x86-64 on a qualified relevant subset of the Proliant range would 'just' be another porting exercise - NSK and VMS have both already been through changes of chip architecture. Such a port would probably be welcomed by the customer base as x86-64 is not under the control of any one vendor and is not going away in the near future. Whether such a port would be welcomed by Big Iron HP, or even practical given the state of HP Engineering ("HP Invent"?), is a different question.

    HP-UX is just another UNIX and there are plenty of those around. One less than there should be though; Compaq and Big Iron HP did once upon a time promise to port Tru64 beyond Alpha.

    VMS and NSK, however, are still each unique in the industry. Some folks are willing to pay a lot for what these OSes offer. Are Big Iron HP willing to ignore these folks completely? It would be a shame (but not a complete surprise) if they were; I'm sure IBM won't mind their business.

    Hurd's role in this looks to be potentially interesting too.

    Where *is* Matt Bryant these days?

  12. Anonymous Coward
    Anonymous Coward

    Correction @What a silly game this turned into

    Correction: The high end Proliant based on Intel x86-64 is now the DL980 G7, 64 cores and 2TB. Still waiting Tim's, Matt's (or anyone's) input on the high end Superdome in comparison.

    And be careful with the HP Big Iron guff about RAS (reliability, availability, serviceability) on IA64 being better too; unless they can come up with specific documented examples, it's just hot air and handwaving.

    1. Steven Jones

      Superdome II

      The Superdome 2 (which is based on a modified form of the C7000 blade chassis used on Proliant) currently scales to 128 cores (256 threads) and 4TB. It has 32 sockets each capable of carrying a 4 core CPU.

      However, I'd certainly agree it isn't going to be in the same price/performance area as a BL980 G7, but the issue for many users of big iron equipment like this is stability and ease of migration. Moving a 10-20TB Oracle DB from HP-UX to OEL is no minor thing, especially if the migration has to be done with only a few hours of downtime and be reversible if it goes wrong.

      Speaking from a company that has a huge amount of Integrity, Proliant and Solaris hardware, then the RAS characteristics of the former have undoubtedly been better than the x86/x64 architecture machines, although the latter is improving through the generations. One issue about very large configurations is that RAS characteristics have to improve as, with many times the components, there are proportionately more things to go wrong. Even things like Oracle RAC won't resolve all those issues as node failures can still leave to brown-outs with stale cache and so on.

      Of course big iron will turn into more and more of a niche, but for some companies the cost of the hardware is sometimes only a modest part of the entire system cost. Stability is often what matters most.

      1. The Great and Powerful Oz

        There are differences

        The DL980 actually borrows technology from the Superdome 2. DL980 doesn't use the "glueless" design of the DL580 or non-HP 4-socket designs. HP found that at 8 sockets, the cache management traffic on the Nehalem Quick Path Interconnects (QPI) consumed a very large part of available bandwidth. They used the XNC (crossbar node controller) ASIC from SD2 in order fix this. The DL980 isolates CPU into 2-socket "islands" that are interconnected via XNCs. The XNCs provide enhanced cache management and fault isolation. That results in much better scalability in the 8 socket system when compared to 8 socket Nehalem designs from other suppliers (including Sun/Oracle and IBM).

        The SD2 still has many advantages in RAS, scalability, and flexibilty. It can support massive I/O. There are a couple of 10Gb/s NICs built into each base blade (8 base blades per chassis). The design supports up to 96 PCIe slots in 8 chassis; the number of slots is not directly dependent upun CPU. The system can be hard-partitioned on cell blade boundaries, and it is possible to remove a blade from service, power it down, and do maintenence without impacting other partitions. X86/X64 systems at present can't do that kind of thing.

        The Itanium processor provides for other RAS features. For instance, there is advanced ECC on PCI I/O paths. You don't get that in Linux hardware, either. Ever see a multi- terrabyte database trashed by corrupted I/O? I have.

        As far as market position, do a little bit of search. It will surprise you. HP is the largest tech company in the world in both revenue and number of systems sold. Bigger than IBM, and FAR bigger than Sun. Most of the volume comes from Proliant, but Itanium outsold SPARC in the last quarter of last year. They're number 2 in the overall UNIX market, but last I checked they had the #1 slot in large UNIX systems. HP-UX owns a HUGE percentage of the high-end SAP market. In fact, SAP runs their internal systems on HP Itanium running DB2. On the X86/X64 side, HP dominates that market.

        Proliant has held the #1 spot in X86/x64 servers for something like 11 years. And HP owns over 50% of the worldwide blade market.

        HP is the biggest dog in the server market. I would hazzard a guess that they won't just walk away from this. It's going to be interesting to watch.

  13. Anonymous Coward
    Anonymous Coward

    @Steven Jones and The Great and Powerful Oz

    Thank you both.

    That paints a picture consistent with what I think see; Proliant is more than acceptable to most of the server market, indeed it is the leader, and outside the chip socket and ignoring the organisational structures within HP, the high end Proliant hardware is increasingly converging with the Superdome hardware.

    More everything in Superdome than Proliant, but for how long, and (relatedly) how many paying customers care? There may still be an odd few details to separate them but is it fair to say the hardware advantages of Itanium are decreasing? PCI ECC? An excellent idea, but surely with PCI now connected via QPI on both x86-64 and IA64, there's no fundamental difference? Or is this a QPI feature not yet implemented on x86-64?

    Is it also fair to say that the distinction between the two families, in general, is in the OS and associated software ? And that the software in question and the IA64 hardware in question both come from HP's Business Critical Systems? Which HP stovepipe owns Proliant? BCS, or elsewhere?

    My home ground OS a few years ago was VMS, though I reluctantly moved on and did a bit of chip/system level stuff before moving on again. On VMS, if you had an interesting business-affecting problem, the in-depth tools and infrastructure (and skilled people) were there to help sort it. Performance monitoring, error logging, crash dump and analysis tools, internals and data structures documentation and expertise, the lot. They may not always be prettified and pointyclicky tools (or experts) but they do what's necessary to maximise availability and keep downtime to a minimum. I assume the same goes for HP-UX and NSK. Wannabee OSes like Linux and Windows, the stuff that runs on Proliants, are toys in comparison, even though the underlying hardware is afaict more than capable of providing excellent RAS.

    Steven feels that his Proliants aren't as robust as his Integrity systems. But is that down to hardware or software or a bit of both? If HP were to put a serious OS on today's high end Proliants (and a few mid range ones too)... then what? And the converse, if/when that doesn't happen, will those customers left high and dry ever buy anything from HP again (Proliants and printers and services included)? If they don't ever buy HP again, will HP even notice?

    "It's going to be interesting to watch"

    Indeed.

This topic is closed for new posts.

Other stories you might like