If you run Itanium-based servers in your data center, 2010 has a surprise for you. The dominant supplier of commercial Linux, Red Hat, is not going to be supporting its future Red Hat Enterprise Linux 6 on any Itanium platforms, old or new. An intrepid reader of El Reg sent us an email saying that some of the Bugzilla reports …
Intel: Put Itanium out of its misery!
Paging Dr Kevorkian!
If Intel wants to compete with IBM's big-iron POWER boxes, it needs to dump Itanium now, and put those resources on Xeon.
Oracle needs to focus on selling Solaris on Xeon. Solaris is the perfect OS for a 32-core, 64-thread, 1/2 TB RAM Nehalem-EX system.
The time has come for Xeon to take on big-iron.
Lack of Integrity
I just love the comment "Intel has been mum on the subject, as has Itanium's biggest cheerleader, Hewlett-Packard."
Yes, I think that'll be a stunned silence from HP HQ in Palo Alto - after all it means that the only non-inhouse OS that HP can run on their Integrity servers is now Windows Server, (the others being HP/sUX and OpenVMS).
If I was IBM I'd be laughing my socks off - especially as HP's been pretty sneering about the merits of Itanic v's Power.
Of course, there's nothing stopping HP from rounding up a few guys in the labs and doing their own distro for the Integrity's. Or - easier still - speaking very quietly to RedHat and bringing RHEL6 support on Integrity in-house to HP. I should think that Mark Hurd wouldn't object at all to the extra services revenue! Heck, it might even mean that a couple more HP engineers that were going to lose their jobs next year get kept on. So there's an upside for someone. :)
In other news...
In other news, bears have been observed defecating in the woods.
Why would anyone who wants Red Hat want it on anything other than a decent industry-standard x86 server, or if the customers are still strategically committed to the mainframe business or committed to PowerPC, on one of those boxes. Why would someone like Red Hat want to continue to support a particular hardware platform if the revenue they get is unlikely to cover the costs of qualification and support for that platform?
For years now there's been almost no interest in Itanium for its own sake, only in the software and hence systems that are still exclusive to Itanium - HP-UX, Tandem NonStop, and VMS, which by coincidence are all HP OSes.
Where the same software is available on a different platform than Itanium, a more mainstream platform whose life support is not in doubt, then users, developers, systems integrators, consultants, etc would be silly to accept Itanium, whereas AMD64 is a perfectly acceptable choice - they're not going away for a while. Meanwhile, mainframes are not going away for a while yet either, much to the surprise of some industry commentators.
There is still a tiny niche that Itanium hardware can address that other non-mainframe boxes can't yet address - the ultra-large-memory single-system-image massive-SMP box.
As the AMD64 offerings encroach further into Itanium "big system" territory, Itanium hardware's exclusive market gets smaller and smaller, and the economics of continued Itanium development in parallel with development of AMD64 chips and systems in the same companies get crazier and crazier too. Intel HQ know this, HP HQ know this, most of the industry knows this (even if they don't admit it in public). The only question: when (not if) Itanium dies, which of the Itanium-specific OSes will HP port to AMD64 (presumably on Proliant, which is already the world's default server for most business-critical IT)? Which of HP-UX, Tandem NonStop, and VMS will HP continue to market post-Itanium?
This Red Hat announcement should surprise no one.
Itanium-dependent customers need a plan of escape, if they haven't already got one.
For clarity: AMD64 here also includes Intel's own last-minute AMD64 clones; the ones that Intel repeatedly said they weren't going to do because IA64 was going to be the "industry standard 64bit server". Right?
A merry Christmas and an Itanium-free New Year to both the readers that got this far.
Of the three to be ported:
between HP-UX, Non-Stop, and VMS, I would expect the first two at least. HP-UX still has a dedicated customer base that HP cannot dump, Non-Stop is STILL a fairly unique platform for ultra-reliability that you could, say, run NASDAQ on. VMS is less clear - a small but dedicated base, but they were never _HP_ customers to start with, they were DEC's. So they may be the odd-ones out.
I remember running a CRM installation for a credit card company a few years ago, and our service provider (Acxiom) kept insisting that we upgrade to Itanium servers - nice and expensively. I had to fight tooth and nail to get AMD64 boxes even considered...because the Itaniums were supposedly "future proof", yessireebob, just like the DEC ALPHAs they were to replace at Acxiom...
Here Matty Matty Matty!
Tell us again about it Mr B, how great is Itanum, how much market share will it have as time passes? Which other CPU architecture will it kill? And quite how fast did your Superdome run RHEL?
More to the point, how great are HP?
SP&L huh? Hee hee....
How many customers get how much of an edge with this combo?
If it's not many and and/or not much look foward to a further Itanium downturn. That's still a fair range of architectures supported and there is a *reasonable* amount of time to prepare. Wheather people will actually start planning or leave it to the last minute is another matter.
Is Itanium shaping up as the i960 for this millenium?
Wonder what Matt B. spin on this will be. He will probably rightly point out that Itanium only there to provide for prior DEC customers that run VMS or HP-UX now anyway. Also would probably also say well see not many Power customers running on RHEL either it will go away soon also. Regardless kind of sad to see Itanium die a slow death ala Sparc and pretty soon x86_64 will be it.
"What IBM needs to do - and should have done a long time ago"
IBM runs scared of the concept of GPL'd stuff. There's always been this underlying fear of this sort of thing that the lawyers have never been able to get over. That was the case when I worked there and it was the case for the immediate years after I left. Until/unless the lawyers get over their fear and/or they work out how to handle GPL'd stuff, this isn't going to happen.
And I've said it before and I'll say it again - IBM's specialists in virtual systems (read the people who come up with zVM these days) can wipe the floor with start-ups like VMWare, Xen and whatever. The reason that there's not a visible product in various hardware product lines is not because the techies can't do it, it's because management don't consider it to be "strategic".
Well, what's the news here?
But if IBM buys Novell it will buy into Novell's deal with MS. That might be a bad taste.
On the other hand, SCO would positively *love* it. They be back in court in an instant arguing that IBM and Novell have been in cahoots all along. And that, since the companies would now be merged, they should have to fight only one lawsuit rather than two.
Silly, twisted fools.
Another IBM Acquisition
Back in the days when Power PC chips were used in the Macintosh, I had thought that - notwithstanding their infamous 1984 advertisement - that IBM really ought to buy Apple. That would have given it a way to make personal computers that had real added value in them, compared to the price-sensitive PC clone market dominated by Microsoft and Intel... which basically turned IBM, the originator of the PC, into just another clone vendor.
Now, of course, it's too late for that, what with the Intel Mac.
I seriously hope,
that RedHat will succeed in making Itanium customers understand that open source is not gratis software else the provider but not the receiver decides it is.
Agree... IBM should have bought apple
They also should follow thru' a lot more.
G5/970? What the hell happened there? That was such a crucial turning point :(
What about the foray into consoles? What happened there with the cell and xenons?
Look, IBM, I can't afford to buy costly power 7 boxen but there are many of us out there who want a viable alternative to x86.
hmm do you really think HP or Intel care about this? Linux on anything other than x86 is and always will be boutique - the difference in actual revenues between a world with RHEL/IA64 and without is tiny.
of course they have to put a spin on it (as the register is doing here with its usual "Itanium is dead" message), but most folks know that IA64 is a play for HP-UX, OpenVMS and NonStop now (with the odd Windows system thrown in - I suspect Windows will stay there whilst Microsoft can only scale-up SQL Server rather than scale-out)
If I were to hazard a guess about the realities behind this I'd posit the following situation:
- HP/Intel & IBM have been providing "marketing funds" to RH for many years (which in fact are payments to port to their architectures
- Either RH asked for more money and HP/Intel said no, or HP/Intel decided to stop the "marketing funds"
- Either way I suspect its only a matter of time before the same thing happens to the Power and Z ports of Linux too.
I had forgotten about VMS and Tandem
Although HP did not ssem to have a problem with killing off the Alpha architecture.
Let's note that if people only buy Itanium because they want VMS and/or Non Stop support *perhaps* HP might consider re-hosting on something more mainstream
After all they don't make the processors, Intel do and having re-hosted VMS from Alpha and Non Stop (I think the Tandem hardware had some proprietary processor) how difficult can it be to find something with more market share where they won't be the sole big design win?
Who should buy what?
--"What IBM needs to do - and should have done a long time ago - is buy Novell"--
Oh I see: after several attempts at trying to convince us that IBM should buy Redhat, now theReg wants to have a pop at Novell. What is this fixation with turning IBM into a Linux distro provider?
--"Acquiring Citrix as well would give IBM an x64 hypervisor (Xen) "--
Yeah right, because Citrix believe in Xen so much that they they put it first above all other hypervisor choices - not (http://www.theregister.co.uk/2009/12/16/citrix_microsoft_site_recovery). Xen has no long term future: KVM, the virtual darling of Linux kernel development will see to that.
HP Itanium market share is tiny
I agree with A. C. #3
HP is the largest x32 & x64 server provider in the market workdwide. Xeon and Pentium familes.
HP-UX/Open VMS on Itanium has tiny market share for 2009/10 new server sales. Outside the USA, where Linux is not suitable for a specific extra-large SMP we see AIX is winning, HP-UX second, and Solaris SPARC nowhere. My worldwide customers ARE asking for Linux on HP Itanium now, but this is only so they can get off HP-UX, convert the hardware to Linux to and simplify their training and supprt costs. Decent sized server buys are Xeons and AMD, and most of those are virtualized.
The interesting next data point will be what will VMware's position be going forward, with say vSphere 5?
The Real Question
Is why does Itanium no longer serve in undermining IBM's competitors? RedHat would not exist if it weren't for IBM you know.
Itanium is dead after Tukwila
Monticello was the only halfway decent chip. Too bad it was 4 years late and one of the last dual core chips in the market.
Montvale the current chip was an embarrassment. It was announced by Intel on Halloween and never really announced by HP. HP did not even bother to update their own SX-2000 chipset to support the faster front side bus.
2008 can and went with no Tukwila. 2009 comes and goes without Tukwila.
1Q 2010 and Intel will finally "announce" the chip, something they already really did in 2007 as "The first 2 Billion Transistor". I guess you dont have to ship something for a few years to claim it being a huge chip.
2Q 2010 HP will finally have Tukwila in systems.
=> Forklift upgrade which makes all purchases today worthless on residual value
=> Twice the performance...oh but wait it has twice the number of cores, so there is ZERO
application performance improvement until Poulson if that chip ever comes to market.
=> Software pricing for Tukwila will increase so customers will actually pay more for less performance. a.k.a. that is a "TCO downgrade"
=> Except for twice the cores of equal performance in the same rack what is the value?
Greetings from McKnight
RE: Here Matty Matty Matty!
Oh dear, it looks like the warning about bitterness didn't get through!
"Tell us again about it Mr B, how great is Itanum...." OK, it's great! Actually, as far as Red Hat are concerned, the problem is hp-ux on Itanium is really great, especially in the enterprise which wants a bullet-proof OS for all those really biz crit tasks. Red Hat have been warning us for a while that this was coming as they are simply not making enough moeny out of RHEL on Itanium (we're a RHEL on Itanium customer, with Oracle on top). As it stands, we have plans in place to move what IA RHEL we have to either hp-ux or to RHEL on ProLiant.
".....how much market share will it have as time passes?...." More and more as the competition is now Power or Itanium. And Itanium scores in the really lucrative high-end, not the webserving arena Niagara plays in. Now, try and concentrate and think really hard about this - which do you think hp wants to sell, an hp-ux support contract or an RHEL one with each server? After all, hp is still the leading Linux server vendor by a country mile, so any RHEL moving off Integrity is going onto ProLiant, not Galaxy or Niagara. To be honest, Dell are more likely to be trusted as a server provider for Linux than Soreacle, and IBM's idea of talking Linux is "sell a mainframe, if you can' t sell that then sell AIX on Power, if that doesn't work then try Linux on Power" - they have no drive or direction for high-end Linux, so I presume IBM are simply paying to keep Linux available for Power and swallowing the loss the usual way - by rooking their mainframe customers another ten points. They managed that for a long time on Xseries before they ceded the field to ProLiant.
".....Which other CPU architecture will it kill?...." Well, SPARC is dead, as is Sun, with Soreacle paying lip service to Niagara. Power looks OK for now, but the roadmap isn't exactly packed.
".....And quite how fast did your Superdome run RHEL?..." With Oracle, very fast, thanks, and much faster than anything Sun or IBM could put up against it at the same price. It was an open tender, all three (plus FSC) were asked to put solutions forward, the vendors were allowed to tweak and tune their offerings as they saw fit, and IA RHEL won. In fact, the shootout really turned into an hp-ux vs RHEL affair as both easily outperformed the competition (only AIX-Power came close), hp-ux was faster but RHEL edged it on costs. I'm pretty confident predicting that if the board decides to make a change then it will just shift to hp-ux for those IA servers we have running RHEL. After all, having bought a standard app stack, it will be relatively simple (and hp effectively already paid for the testing in the shootout!).
"......More to the point, how great are HP?...." Hmmm, let's see - still number one Windows server vendor, ditto Linux vendor, and still number one high-end UNIX server vendor. I won't bother with the other areas (storage, print, networking, etc) as I know that reminding you Sunshiners of the narrowness that killed Sun just makes you froth all the more, and I really wouldn't want you crying at Christmas. I just hope Santa brings you a new hp printer for Xmas, as that will really get you squealing!
".....SP&L huh?...." Well, it's hard not to when you Sunshiners are so comic! So, how many times is it you guys have predicted the end of Itanium? Must be at least every quarter for the last nine years! Coincidentally, about the same period as it has been since Sun made a profit - not much chance of that now. Before you get too excited, consider a few things - firstly, hp have been driving the IA version of Linux for years, so - if the feel the need - they could push another distie, such as Oracle's own; secondly, Linux is still more strategic to Oracle than Slowaris, even with Sun's laughable Linux-clone Slowaris x86; thirdly, RH are Larry's enemies, not firends, whilst he still sells five times as many Oracle licenses on new hp servers than Sun kit, even if only on ProLiants, which means hp are his bestest buddies (ever wonder why all Larry's public attacks are on IBM?). Oh, I thik there are still plenty of other people pointing and laughing at you guys, and most of them probably work at hp. As it is, my own arm is starting to ache as you guys have given so many years of hilarity - haven't you all dried up and blown away yet?
/SP&L, with a Merry Christmas to all!
Tsk tsk Matty!! 4409 words as a reply, your rants reveal when your rattled but hey ho, you look bored quite often. Being a HP ad man must be fairly boring I guess, not much innovation to write about.
However it's more worrying that you look as though the senility is starting to take hold here, you appear to have forgotten Sparc. Lets help seperate reality from Matty world here :
In Matty world all that exists is HP versus the world and HP are the leader, Sparc is dead and long live Mr Hurd! (sound of crisp salutation and heels stamping!)
In the other world, ie: the real world it appears to be sold by Sun and Fujitsu with a promise of more to come from Mr Ellison. Sales hesitation will occur in the market place due to merger fears but you and I both know whats comes once approved. The axe falls, dead wood is chopped, sold or simply burned and the stronger products step forward with much more cash behind them. You won't like admitting it but it doesn't look like Sun will be losing more market share in 2010 once Oracle owned.
Let's stop the Sun talk Matt and chat about the little point of Linux being firmly embedded on x64, Solaris x64 being accepted since the Open Solaris re-awekening and yet HP-UX and VMS are on a promise not to be ported across to new architecture, again??
How many ISV's will totally drop HPUX / VMS support at that point do you think?
How many customers will drop HP OS's at that point and just go x64 on anyone's tin?
Hmm, more fish I suggest Matt, heard the pottasium content helps the brain cells along. Anyways, gotta dash, expecting some vehement rhetoric any moment soon.....
This is going to impact SGI too as RHEL AS is one of the two OS options (the other being SLES) for their Itanic-based Altix 4x00 systems. And if Novell decide to drop SLES on Itanium too...
Serves them right for not porting IRIX to Itanium, I say ;-)
SGI is not going to support Tukwila either...nor Unisys
The writing has been on the wall for a year. Itanium = Tandem and HP-UX
If Xeon could switch it's Endianness Itanium would be dead today.
- Vid Hubble 'scope scans 200,000 ton CHUNKY CRUMBLE ENIGMA
- Google offers up its own Googlers in cloud channel chumship trawl
- Bugger the jetpack, where's my 21st-century Psion?
- Interview Global Warming IS REAL, argues sceptic mathematician - it just isn't THERMAGEDDON
- Apple to grieving sons: NO, you cannot have access to your dead mum's iPad