1748034 Members
4812 Online
108757 Solutions
New Discussion юеВ

Re: tukwila

 
ABrown_1
New Member

Re: tukwila

I believe that there is a serious issue with the current Montvale servers. No future proofing and so expensive. Tukwila will make a huge difference in the amount of CPU power in a smaller chasis with better power management. Who here can remember the change to PARISC when it went to Dual Core back in the early 2000's? The current server range are a good set of over priced boat anchors. Tukwila will have to compete with power7 (IBM are offering inbox upgrades) once eventually released. My recommendation to all is to get HP to lend you HW until Tukwila is released. Tell them you are not interested in buying legacy equipment at crazy prices. Bring on HPUX on Nehalem, that way us HPUX experts may still have some good future job prospects.

Re: tukwila

>> Tukwila will have to compete with power7 (IBM are offering inbox upgrades) once eventually released

>> My recommendation to all is to get HP to lend you HW until Tukwila is released. Tell them you are not interested in buying legacy equipment at unreasonable prices. Bring on HPUX on Nehalem, that way us HPUX experts may still have some good future job prospects.

Well now I work for HP, so I have a vested interest, but I'd make a few points here (and these are *my* opinions, not HP's) -

i) IBM boxes required a box-swap for every upgrade until the p5 to p6 transition, (and even then I'd advise you to look at the small-print of what an inbox upgrade actually means...) - plenty of folks still bought these systems knowing they'd have to move to new ones for the next processor rev... The basic architecture for HP-UX systems, both low-end and mid-range hasn't chnaged much except for speed bumps in 10 years - thats why we've been able to offer in-box upgrades for so long. Now we need a new architecture for the next 10 years - you have to do it sometime - this is technology after all and these things happen - go down to your local auto sales yard and tell em to loan you a car until the new model comes out next year - I expect you may get some odd looks!

ii) Unreasonable prices? Maybe you need to ask for a bigger discount? In my experience we're generaly cost competitive with AIX and Solaris systems - Not with x86/Linux systems but I'd argue neither are Solaris/AIX and you need to look at TCO and not just acquisition costs

iii) HPUX on x86-64? I don't think so... why would HP do that? Competing with AIX and Solaris is OK, but going up against Linux and Windows in the x86 space would be a *huge* waste of time and money. I'm not saying I wouldn't _like_ to see it, just I can't see it happening. Just look at the trouble Sun have had getting any sort of market acceptance for Solaris on x86. Sorry but UNIX on x86 = Linux.

HTH

Duncan


I am an HPE Employee
Accept or Kudo
Michael Leu
Honored Contributor

Re: tukwila

Thanks for your opinion Duncan.

By the way I just went this route by lending two systems until tukwila will arrive.

I think the hardware price itself is competitive, but IMHO the per core licensing and support of HPUX makes the system very expensive again. I pray they change the license model with the upcoming quad-cores...

And while HPUX on x86 would be nice, I agree it probably wouldn't last long (just look at ServiceGuard for Linux).
ABrown_1
New Member

Re: tukwila

Im sorry but recent numbers do not lie:
"Cathie Lesjak, HP's chief financial officer, said the new ProLiant G6 servers, which used the latest Intel Nehalem EP Xeon 5500 and AMD Istanbul Opteron 2400 and 8400 processors, helped push revenues up 14 per cent compared to the second fiscal quarter." whilst ....
The Business Critical Systems unit, which is where those Itanium servers and some ancient parts for PA-RISC and Alpha systems come from, declined by staggering 30.3 per cent, to $578m.

Re: tukwila

You are correct - numbers do not lie - BUT they do need to be looked at in context.. if you look for instance at recent industry figures (I won't cite refernces as you didn't) you will see that in the last quarter the whole UNIX (RISC/EPIC) market shrunk by about 30% - pretty much where BCS numbers are - that's what recessions do... In the same quarter, Sun revenues for theirs SPARC systems were down 40%! I'm glad that in these tough times, HP has the cusion of its volume ISS business which is looking to have made a recovery (didn't looks so good a few quarters ago) that neither IBM or Sun have...

Again my opinions, not my employers and I don't make any claim to not being biased - just wanted to throw some context out there for your numbers

HTH

Duncan

I am an HPE Employee
Accept or Kudo
smatador
Honored Contributor

Re: tukwila

Hi,
For Duncan, when you write UNIX on x86 = Linux for me is like writting UNIX is dead. Why? Let me explain what's happen with my customer. For him, the price of hpux and aix is too expensive { all customer said that }, so they decide for example to put all new projects with database on linux. So they'll plan to put an oracle database of 20TB on linux with the next BL680c G6. So, finally if they succeed what could be the future of UNIX? If Linux is competitive in performance, comparing with the price of UNIX, I really think the war is lost even with tukwila.
Regards,

Re: tukwila

smatador, If that sort of argument held water, there would be no IBM Mainframes out there would there? Last time I checked IBM still make _huge_ amounts of money from their System z business...

The UNIX vs Linux argument is a whole different one from the points I was addressing above. No one in their right mind will dispute that Linux has lower acquisition costs than any commercial UNIX you might mention (until you need to scale single workloads beyond 4 sockets...), but surely you realise that acquisition costs are just one part of the equation?

TCO is the key to a platforms position in a market, and whilst you can't as easily put your finger on TCO as you can acquisition costs, it seems to me enough people still see the "value" in running commercial UNIX products over Linux. 10 years from now? I don't know, but I do know Linux right now is still missing too many features that my HP-UX customers take for granted.

Duncan

I am an HPE Employee
Accept or Kudo
smatador
Honored Contributor

Re: tukwila

Hi

I agree with you that TCO is the key to a platforms position in a market.
On HP website, I read this TCO study
http://h71028.www7.hp.com/hpux11i/downloads/hp-ux_11i_roi_aix.pdf
So for IT Cost and direct cost linux x86 win for hw, sw, it operation etc, not really sure for indirect cost but linux improve every day its features, so what linux is missing now compare to HPUX?

LVM is developp by HP. Mondo by HP.
Even the multipath is improved.

Look at this suse presentation... novell is comparing tco solaris to suse, ok solaris is not hpux, but for how many time...
http://www.blu.org/old/meetings/2009/02/fcurran-tco-presentation.pptxv1BLU_GH_02-14-09final.pdf

Re: tukwila

>> so what linux is missing now compare to HPUX?

Well off the top of my head my own personal list would have include:

- Online add/replace/delete of hardware components (CPU, memory IO cards) - linux has some of this but it is generally pretty weak.

- Utility pricing capabilities (like iCAP, TiCAP, GiCAP)

- Some sort of virtualisation capability without the virtualisation overhead (like vPars)

- funny you mention MPIO - a _standard_ approach would be nice instead of dozens of different implementations. Does Linux still have that horrible habit of moving my disk and LAN IO device files around as well?

- Proven vertical scalability (show me customers running Linux on systems with > 4 sockets - I know the kernel is capable of this, but the proof points are few and far between)

- A decent set of workload management tools like PRM, WLM and gWLM.

- A _consistent_ approach to management - one of the things that drives me mad about Linux, is if you give 10 Linux admins a sysadmin task to complete you'll get 10 wildly different solutions - this makes many Linux implementations very "bespoke" and its hard to just pick an admin up off the street who can run these bespoke systems - yeah this ones pretty subjective, but thats what I've seen.

- Less change! The pace of change in Linux is too quick for most enterprises - this means they often find their ISVs have certified products against different OS releases and as a result it can sometimes be difficult to create a standard OS stack across your enterprise. Add to this the pace of hardware change in x86 land with its 3 year refresh cycles and things are even worse.

HTH

Duncan

I am an HPE Employee
Accept or Kudo
Wim Rombauts
Honored Contributor

Re: tukwila

I came across this thread in my search for Tukwila and new HP hardware.

I last heard that Tukwila should come to market Q1 2010. New HP hardware should start coming april/may.

But it remains so frightfully silent about all this. It won't be the first time that intel delays the processor at the very last moment. Performance data is vague (2x current montvale). And where does that put tukwila against power6 and power7 ?

So I make this post here to see if any of you has more recent or more complete data.