Slashdot Mirror
HP Sues Oracle For Dropping Itanium Support
Fudge Factor 3000 writes "HP is suing Oracle for a breach of contract, claiming that Oracle was contractually obliged to continue supporting the Itanium architecture, which they recently nixed support for. Oracle has fired back that Itanium is essentially a dead architecture and will soon be discontinued by Intel. And so the blood feud continues between Oracle and HP."
Why in my day Oracle had to support my UNIVAC for fifteen miles in the snow barefoot uphill both ways!
and any other company following this issue is that they're essentially at the mercy of the business decisions of a third company, Intel, and that's not a very smart business position to get in in the first place.
"A door is what a dog is perpetually on the wrong side of" - Ogden Nash
http://www.theregister.co.uk/1999/10/29/intel_execs_outline_y2k_chip/
12 years of sarcasm!
Makes sense. HP probably has several contractual obligations of their own since there are many large corporate clusters which are using HP-UX running on a shitload of Itanium systems.
You are the epitome of modern corporate culture. You destroyed the Alpha and are letting VMS rot. You outsource or offshore everything that isn't bolted down, but nothing is improved. Under Fiorina you demonstrated precisely how to run a company down for short term profit while cosying up to the corporation-friendly government. Hell, you've even ruined your reputation for building hardy calculators. Over a decade after this mess started, the only thing you have left to be proud of is the propotion of your profits which come from selling printer ink.
It's a small wonder zombie Hewlett and Packard haven't risen from the grave, given a new lease of life in death by recently shuffled Olsen, to personally escort every HP executive to the lowest region of hell.
Oracle should rename mySQL as "Oracle for Itanium" and send it to HP.
> . this means one of two things - Oracle are lying (believable)
Oracle are lying about what? There can't be more than a few hundred Itanium users around the world, and Intel has been signaling for years that the product line is dead and won't be replaced. Maybe HP shouldn't have shut down its own CPU development and sold its designs (along with the designs of the DEC Alpha) to Intel, but they did. Now that product line has failed and Oracle is just making an obvious business decision.
sPh
I think the bigger question is this....why is Intel continuing to beat the obviously VERY dead horse that is Itanic? Its a giant flop, YOU know this, I know this, and Oracle knows this as well, so why continue to waste R&D for a chip that barely has even a teeny tiny niche and is being phased out by almost everyone?
The problem with Itanium is that Intel bet they could not only get everyone to abandon literally billions of lines of already paid for X86 code, but that they could build a compiler able to keep it fed and do all the heavy lifting and in the end they just weren't able to deliver compared to X86-64. Like it or not for many jobs X86 will be here to stay for a long time and Itanium was never a real contender.
So why are they wasting their money? It isn't like they don't have a very valuable product line to replace it, where money is no object Xeon rules the roost in performance by a pretty big margin in servers, just as in the desktop for sheer power the i series owns the top end (for the rest of us Opteron and Phenom work just fine, thanks) so what is the point? it isn't like they are gonna magically get everyone to suddenly drop X86-64, POWER, and Sparc, all of which are beating the Itanic, and just the fact that I can say itanic and everyone knows what I mean just shows the chip has a bad rep. Let it die already Intel, throw HP a sweetheart deal to say you're sorry for the oopsie and just let the thing die already.
ACs don't waste your time replying, your posts are never seen by me.
If Oracle wants to countersue HP for dropping the Alpha, I'd be happy to testify on their behalf. The idiots at HP killed off one of the best architectures for scientific computations ever.
Damn_registrars has no butt-hole. Damn_registrars has no use for a butt-hole.
Itanium isn't widely used; but its users would appear to punch well above their weight when it comes to willingness to pay. Apparently, Intel does ~4 billion a year in sales of the thing. Now, I suspect that continued development is costly enough that Intel would be much happier if they could convince everybody that "Xeon; but with the RAS features not lasered out and for 20x the price!!!" is the way to go; but the sort of people who spend huge amounts, per unit performance, on fairly obscure architectures are exactly the sort of people who Do. Not. Want. to switch architectures.
Intel and its remaining Itanium buddies(yeah, HP, that pretty much means you...) are in the unpleasant situation of having a product line that is too big to just drop; but likely much more expensive than they would like to keep remotely near performance parity with x86 and with a customer base that is unlikely to just quietly accept an architecture switch in the near future. HP's situation is, of course, rather worse than Intel's... since whenever the ship finally sinks, so do HP's remaining offerings in the "architecturally better than boring x86" niche. IBM will still have POWER, Oracle will have at least the option of carrying SPARC forward, Dell will still have rock-bottom prices...
I doubt they are doing too much R&D on it, they are mainly just manufacturing the CPUs(in small quantities I'm sure) so they don't anger existing customers. They only stopped making 486s in 2007(those most of those were for embedded applications)
Also, just my 2 cents, Itanium didn't fail because compilers couldn't effectively utilize it, it failed because VLIW was an academic experiment that got waaaaaaaaaay out of hand. While compilers certainly could have utilized it better, they cannot violate the fundamental constraint of ensuring at COMPILE TIME that no 2 instructions being executed simultaneously have any data or execution dependencies. Compare that with the superscalar design used by most CPUs today: they can use runtime behavior to predict jumps, data dependencies etc. While backing out an instruction that was partially executed is costly, modern superscalar designs have to do it so rarely that the little bit of performance penalty for such instructions pales in comparison with the gains you get when you execute multiple instructions simultaneously that *might* have a dependency, something VLIW simply cannot do.
Monstar L
IANAL but it seems to me that if you have a contract with someone that says you are going to support their hardware they have already purchased, then it is irrelevant that the hardware vendor *might* be discontinuing the product. Even *IF* Intel discontinues it, Oracle should still be bound by their support contracts.
and Oracle is just making an obvious business decision
But (assuming HP are correct about the contract terms) that decision is in breach of what they have originally agreed.
If HP aren't blowing hot air, then it will cost Oracle. Though they may have planned for this eventuality and calculated that "sum(max(amount_could_sue-for+legal_expenses)) from clients where contract_says_architecture_will_be_supported_longer=1 and likely_to_sue=1" is smaller than the cost of continue support.
And where do contractual obligations fit into all of this? Rather than being sued, I wonder if Oracle should not have negotiated out of their contract with HP. It's a funny and hypocritical world of business we see these days. Copyright organizations ignoring copyrights and violating license terms they seek to enforce against others. Companies which famously viciously defend their own trademarks while trampling on those owned by others. Contracts, and other business agreements simply disregarded likely because they think they can get away with it.
I know it's not new to describe corporate behavior as sociopathic, but it is and the problem seems to be getting worse.
We like to say "just business" but there is no "just business." To paint it grey in order to cover up the obvious lack of good faith, ethical and moral character erodes humanity in ways I cannot begin to describe. But taking that route, couldn't we also characterize mass killings and murder in different parts of the world as "just business" too? Or how about mass starvations and death by preventable/curable diseases at the hands of people who control the intellectual property of these things. How about putting people into homes under terms they cannot afford and creating securities with the loans which result in serious financial problems for the planet which result in all sorts of tragedies the world over? Still "just business?"
"Global leadership" is more than government leadership. It is also business leadership. And so far, there are seemingly too few controls against that leadership and the damage they can do to the rest of the world. Ignoring contractual obligations is merely a small sign of a much larger problem. We mere mortals are not allowed to ignore contractual obligations and neither should mega-business.
HP was a partner since the mid 1990's when development started and i bet there was a contract for a minimum life since HP also invested a lot of money. back then it was very different where HP and Intel made low end products and were salivating at the thought of selling a competitor to SPARC and similar products with their insane margins
There's nothing wrong with SPARC. Solaris/SPARC is still powerful, innovative, scalable and secure. AIX/Power would of been a good match, but Big Blue doesn't invest in AIX. Solaris had more new features added in the past 5 years than AIX in the past 15. SPARC is still going strong because Sun realized that it's not the GHz but the 6,8,16 cores/die that are the requirements for a server and because Solaris is still going strong. The Sun Fire T1000 is still unbeatable in cost/performance even if it's 5 years old.
All the Linux vendors (HP, Dell, IBM) keep trying to migrate users from Solaris to Linux claiming that it's dead. Yet, they are wrong. HP-UX, Unixware, AIX, Tru64 are however dead or dying.
Think about Solaris the following way: Oracle Exadata 2. Still nothing on the face of earth like it, because of Solaris and Sun hardware. SPARC is going to do even better once Oracle ports it's Linux distribution to it. Since Oracle Enterprise Linux and Oracle Solaris are eventually going to be the only operating systems certified for Oracle Software, you're going to see a lot of them, especially on SPARC. Oracle, after the acquisition of Sun is slowly changing into an appliance vendor, and I can't blame them for it.
UNIX was not designed to stop you from doing stupid things, because that would also stop you from doing clever ones.
Oh please, powerful people have acted like this for thousands of years, it's not the world of business "these days" nor there's anything to "erode."
Not that it makes it right, but it's nothing new.
Dilbert RSS feed
Back then, HP owned PA RISC and Alpha. They weren't 'salivating at the thought of selling a competitor to SPARC,' they owned the chip that was the undisputed performance king. What they wanted was to outsource their chip R&D and production to Intel, without losing their market lead. They stopped developing chips in house, and sent their chip designers over to work with Intel. Now they're stuck with a couple of operating systems that only run on overpriced chips.
I am TheRaven on Soylent News
Intel does ~4 billion a year in sales of the thing
Have you seen the price of Itanium kit? That's about six customers...
I am TheRaven on Soylent News
I confirm, you are not a lawyer. Breaching contracts and having our clients let go off scott free is the basis of our business.
because itanium reaps 4 billion a year for intel. itanium alone brings more revenue than all AMD products put togheter.
from arstechnica
What ? Me, worry ?
not just powerful people. plenty of joe sixpacks do sleazy stuff too. the only difference is the amount of resources involved.
If Itanium is dead, then why does Intel have all this architectural investment?
And? What's the opportunity cost of $4b of Itanium sales? If the opportunity cost is $4b of x86 sales, then what's the point of continuing with the Itanium farce?
(I imagine that $4b of revenue from x86 sales is more profitable than $4b of revenue from Itanium sales - because the Itanium-specific fixed expenses are amortised over a tiny number of Itaniums produced per year, leading to high fixed costs per processor, compared to the x86-specific fixed expenses being expensed over the massive amount of x86 processors Intel churns out, leading to low fixed costs per processor).
The article is worded incredibly poorly in order to try to force a point that doesn't make sense.
It compares the $4 billion per year Itanium revenue stating it's higher than AMD's combined $1.6 billion for Q1 2011. It makes absolutely no sense to compare one yearly revenue figure with another quarterly one.
If that be the case my corner gas station likely makes more money per year *than all of Microsoft combined* (between 3:01:31AM and 3:01:35AM on August 4).
To compare more accurately, AMD's 2010 revenue was $6.5 billion, which is indeed greater than Intel's Itanium revenue of $4 billion over that same period.
"People who think they know everything are very annoying to those of us who do."-Mark Twain
For everytime I had to call for support, or try to use their itrc, or everytime I had to come up with a model number from a device that had 20 different numbers on it, none of them matching the format they expected, for everytime I had to deal with HPUX...serves them right.
The Mongolian cluster-fuck that the Intel-HP-Oracle triangle centered around Itanium has become, was spotted way way before any of their management drank the Kool-Aid. (And oh man, they were guzzling the Itanium Kool-Aid by the gallon). Google "Itanic".
The Itanic failed because it was more expensive than the x86, not faster in "integer" performance and not compatible with the x86 for practical purposes.
;).
:).
Back then (and even now) if you had lots of money to spend, wanted "Itanic" style performance, and didn't need x86 compatibility, you might as well buy an IBM POWER system. At least you know IBM will be happy to keep sucking money out of you for decades
The rest of the world was and is better off using x86 systems for their servers. Most of the stuff the Itanic is good at is easily run in parallel, and so can be run on multiple cheaper x86 machines, or nowadays on multiple GPU cores
The x86 on the other hand can do the usual web/DB/"misc app" stuff faster than Itanium. The x86 specint scores were better, while the itanium specfp scores were better. The last I checked, stuff like Oracle, Apache, IIS didn't require much floating point performance. If you're willing to dedicate the chip space and power budget, I'm sure it is possible to make x86 floating point performance much faster, but who will pay more for that?
If that be the case my corner gas station likely makes more money per year *than all of Microsoft combined* (between 3:01:31AM and 3:01:35AM on August 4).
Except the original statement in the article is probably true, where as there is no way that gas station makes more money in several years than MS does in any given second on interest alone. You seriously underestimate the amount of money they have sitting around.
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
Except the original statement in the article is probably true, where as there is no way that gas station makes more money in several years than MS does in any given second on interest alone. You seriously underestimate the amount of money they have sitting around.
Fortunately, this is a simple math problem.
365 days per year * 24 hours per day * 60 minutes per hour * 60 seconds per minute = 31,536,000 seconds per year
Microsoft's yearly revenue is between $65 and $70 billion. We'll take 2010's numbers of $66.7 billion. That equates to only $2,115 per second. The original statement was a 4 second span - we're still talking less than $10,000, which a big gas station can easily take in in a week or less.
"People who think they know everything are very annoying to those of us who do."-Mark Twain
So, do they have a contractual obligation to keep the port going or not? Whether it's a dead architecture doesn't matter if they took the money and there's no 'dead architecture' clause in the contract.
Hey, Oracle guys: talk to the Redhat Itanium team. Last I heard they were passing the hat around the office and were going to buy the remaining few Itanium machines left in the world and throw them off the roof at HQ (and then promptly recycle the remains, I'm sure).
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Intel makes more money on Itanium than AMD makes period...
I'm using all of my mod points to mod ancient memes down. Please join me.
The same could easily be said about the sparc architecture.
but the chip being sold as "Itanium" is missing some of the original claimed core capabilities, Oracle could easily prove "this isn't the Itanium we were contracting for"
even though down for the last quarter, Oracle's sparc server sales were over 70% of the RISC Unix server market for 2011 thus far. not dead yet.
The irony of Itanium is that while still at Intel, Andy Glew (the Pentium Pro architect) proposed a 64-bit architecture very similar to what AMD eventually did with their AMD64. Obviously Intel rejected the proposal because their heads were so far up their VLIW asses.
> why is Intel continuing to beat the obviously VERY dead horse that is Itanic? Its a giant flop,
When you look at it from a Marketing point of view, Itanium was perhaps the most successful chip in history -- not because it sold well, but because it bought time for x86 to move into 64-bit computing.
Recall that a few years back there were a number of promising 64-bit architectures that were a lot cleaner than Itanium. HP's PA-RISC and SGI's MIPS were two of them. Yet even before production Itanium chips launched, the CEO of SGI and the Executive VP of computers at HP made the decisions to move away from HPUX-on-PA-RISC and IRIX-on-MIPS in favor of NT on Itanium, pretty much killing their entire computing product lines, and killing R&D on the server versions of both chip families.
Seems the guy (yeah, same guy) who made those decisions was rewarded with a President and COO job at Microsoft a bit later.
Intel may not have won with actual sales of Itanium hardware (and I wonder if it even wanted to - from it's point of view X86 was even better because HP had some joint-ownership-rights in Itanium). But it sure won by killing the competitions 64-bit computing roadmaps from the PR of Itanium alone.
http://quiller8771.blog.com/2011/05/26/double-s-fall-it-technology-in-the-market/
http://www.theregister.co.uk/2006/05/13/sgi_belluzzo/
HP has to beg a single vendor to continue producing wares for it's dying chip, because that is the only remaining justification for producing servers based on Itanium. Give it up already HP. Oracle has no such contractual obligation.
How is it a giant flop? It passed SPARC sales years ago. VMS, HP-UX, and NSK all have thousands of customers with no obvious migration path and billions invested in the platform. For these customers, Itanium works fine, and has certain advantages (socket compatibility over multiple generations, for instance) and no obvious reason to dump it.
On the other hand, you totally lost credibility when you said SPARC was "beating the Itanic." (looking up sales or performance figures from the last 3-4 years would be educational.)
"A few hundred?" Tens of thousands, minimum. NSK and VMS alone have a few thousand users each, and HP-UX is much bigger than either. Learn about the industry before sounding off.
I'm curious, have you seen any of the SPEC_cpu benchmarks on SPARC64? You know, the ones that put a quad SPARC64 VII+ at performance roughly equivalent to a dual-core Phenom that's ~99% cheaper? How about the ones that show no marginal advantage of a SPARC T3 over a Magny-Cours Opteron, even when continiously saturated with threads, again with a vastly cheaper acquisition cost?
Didn't think so...
http://www.spec.org/cpu2006/results/res2011q2/
Er, no they weren't. Oracle's total (x64 and SPARC) sales for Q1 were $798mn. The total RISC-UNIX market for Q1 was about $2.6bn. IBM's POWER sales were $1.2bn. http://www.theregister.co.uk/2011/05/26/gartner_q1_2011_server_numbers/
I'm curious, are you on the Oracle payroll?
no, not on payroll and I despise Oracle and it's bloated 1980s technology DBMS, and also what Sun had degenerated into since the dot-com bubble when they became careless and just rode out the name. Doesn't change that UltraSparc is still heavily present in datacenter and still purchased even in recession. Gartner is the National Enquirer of the IT world, they don't know fact from marketing fiction. Stats on sales are a function of who's reporting and how they want to count things as bundled ( for example server plus storage plus software if you want to jack up numbers)
It's been widely known for years that IBM Power is on top by quite a bit, with Itanium in distant second and SPARC occasionally catching up. SPARC has been fading for years though, as Oracle's own revenue figures strongly imply.
I don't care if they contacted to support an Atari 800 for 40 more years. If they breach contract they are liable for damages. That the Itanium is going away is not relevant to the discussion at all. They agreed to do it, so they have to. ( unless there was some special clause in there tied to availability from Intel, which it sounds like there was not )
---- Booth was a patriot ----
Apollo Computer was shipping the 64-bit PRISM workstation when they were bought by HP. HP killed the PRISM because they were going to do their own 64-bit architecture.
Digital was shipping the 64 bit Alpha machine when they were bought by Compaq which was then bought by HP. HP killed the Alpha because they were committing to Intel's Itanium.
So what happens? HP, the owner of two, market-proven, debugged and viable 64 bit architectures finds itself backing the loser, having killed both of the projects they bought and paid for.
And so, what is HP's 64 bit architecture in the end? The x86_64.
You've really got to wonder what kind of idiots were running the company.
Intel does ~4 billion a year in sales of the thing
Have you seen the price of Itanium kit? That's about six customers...
Have you seen how much those six customers pay? That's about 4 billion dollars...
Are you adequate?
Intel is still working on Itanium and has a new architecture in the pipeline. This article agrees that simple VLIW is dead, and indicates that the new Itanium architecture will do a little bit of scheduling on its own instead of relying completely on the compiler.
But can you dance to it?
It's people like you that are the reason the world hates America. Take your racist bigotry, and piss off.
For a site about things like basic rights, Slashdot users sure do like to censor "dissent".
It was under them that support for NT was dropped, and once that happened, the architecture was dead: people were not going to prefer OSF/1 to HP/UX or AIX, let alone Solaris or Linux.
I never really understood why anyone wanted NT for the Alpha anyways. There was so little of anything useful that could be done with NT back in the day, in comparison to what you could do in *nix with an Alpha. HP/UX was great on the Alpha, although we eventually setup Linux on ours to make it easier to install binaries for new software.
In the end I just don't see the Alpha making sense for a Windows user, any more than I see Windows making sense as a serious computational environment - which is where the Alpha was best suited.
Damn_registrars has no butt-hole. Damn_registrars has no use for a butt-hole.
> The problem with Itanium is that Intel bet they could not only get everyone to abandon literally billions of lines of already paid for X86 code, but that they could build a compiler able to keep it fed and do all the heavy lifting and in the end they just weren't able to deliver compared to X86-64.
The problem with VLIW (and RISC) is that the compiler has to optimise the code for the CPU architecture, but the CPU architecture changes from generation to generation. You optimise for one chip, and it will run slow on the next generation, and not at all on the previous. SPARC had that problem, but they managed to produce an excellent compiler with lots of switches for tuning the compromise. But software developers don't like this kind of inevitable incompatibility, and they much prefer the much more gentle progress of x86.
Fuck off, APK. Your hosts file stuff I can live with, but I draw the line at racist trolling you piece of crap.
(Note: not Jewish, or American).
For a site about things like basic rights, Slashdot users sure do like to censor "dissent".