HP Integrity server family Bad Rap

[Matthew Ghofrani]

I am constantly being told that HP's Itanium servers with Intel chips are really horrible and HP has made a big mistake going that route.

I really want to know your honest to goodness impression, whether or not you have had first hand experience and whether or not you had done any comparison with others.

We are a big rp8400 house and I have to be honest it bothers me if I find out the stories I hear are all made up for one reason or other.

Please be candid with me and share your thoughts.

Matthew from Boston

[Pete Randall]

Matthew,

I believe that most of the negative reactions came from the earlier generations of Itanium, which did have some problems. The latest versions seem to be just fine.


Pete (from Vermont)

[Arunvijai_4]

Hi Matthew,

As pete said, Itanium 1 processor family was having problems which were rectified by newer generation Itanium-2 processors. Itanium 2 is performing well on mission critical platforms and it will be the future of 64 bit computing.

-Arun

[Rick Garland]

We have numerous rx4640 systems with HPUX 11.23 and they are running MCSG 11.16 & 11.17 with Oracle 11.5.10

We have an uptime of 99.97% for the last 10 months on these rx systems. They have been solid performers. These are connected to 2 redundant fibers with SAN storage on 2 different EVAs. (EVA 5000 and EVA 8000)
These systems are CSS systems so we keep the patches up to date. We have semi-regular maintenance windows (sometimes do not use so these systems are not rebooted on a regular basis). Performance has been good.

Doing some tasks requires a different methodology. Examples, mirror the root disk, booting from an Ignite server, the EFI, etc.

We have not been doing any partitioning or running any other OS besides HPUX so I have no knowledge there.

But for HPUX, been a good system.

[Matthew Ghofrani]

Thanks for sharing your inputs. Any other stories good or bad? I am really interested to know about where these bad rap stories are coming from. I am not convinced that just having some issues with Rev 1 would have created the enormous negative wibes that I get in hour rp8400 house. I have recently read that HP and Intel and bunch of other companies have pledged a 10 Billion dollar investment toward Itanium chips and for the life of me I just can't buy the idea that they are all bunch of stupid people which can't see the light. Meaning they are heading down a wrong path. I really appreciate knowing more of what you think specially folks if any which originally wanted to go with Itanium but changed their mind or vise versa.

Matthew from Boston

[Torsten.]

Hi Matthew,

this question first: Are you satisfied with the rp8400s? At the moment, all PA-RISC systems have their equivalent Integrity system. For example rp8420 and rx8620, rp3440 and rx2620, rp4440 and rx4640. The same hardware (more or less), you can run the same OS (11i v2 for the moment), but they have different CPUs. In the Superdome, you can even have PA-RISC and Itanium CPUs at the same time (in different partitions).
Newer servers in future will have only Itanium CPUs, I guess. They are top rated in different performance measures.

The biggest problem is the public image that the name "Itanium" has. IMHO, it came from the first generation in conjunction with a Windows XP like OS version. At this time, the CPU was big, expensive, power consuming and *slow*. It could be compared (with this XP-like OS) to a normal PC and the PC was faster and cheaper.

A PC couldn't have up to 64 CPUs or more, but this was never the point. In this time, the first hp Itanium system came out (something like rx96xx), but I've never seen it.

The bad start of this cpu is the biggest problem.

Nowadays the Integrity systems are very reliable and fast. This is my experience.

Regards
Torsten from germany

[Matthew Ghofrani]

Thanks Torsten from germany for your input.

[Matthew Ghofrani]

I am personally satisfied with RP8400s. We have 6 of them with one housing an SEU. We have a total of 20 servers in this area with mixture of physical and virtual servers while running variety of apps like Oracle, Peplesoft and others in a MC ServiceGuard env with about 8 nodes and just about 10 packages.

Our move is toward more virtualization and more bang for the bucks that Itanium does offer along with other competitors.

That's were I come in to really get the myth out of bad or good perceptions out of different hardware/platforms.

Matthew from Boston

[Torsten.]

The RX can do what the RP can do for you and even more. With new CPU generations and the sx2000 chipset the Integrities will become faster and faster. You can run different OS versions like HP-UX, Linux, different Windows and OpenVMS. Even NonStop servers are using Itanium CPUs now.

In terms of virtualization the vPar solution is working fine, but virtual machines (only on Integrity) will provide much more flexibility.

Remember, there are a couple of millions PC "experts" talking about Itanium, but only some thousand people have really seen it working.

[Paul Jerrom]

Howdy,
Let's face it, FUD (Fear, Uncertainty, Doubt) will be thrown by all competitors and luddites.
We trialled some early generation Itanium servers, and were not impressed by the performance. However, the early hardware was running early operating system and early layered products - and the behaviour on Alpha is different to that on Itanium.
We now have low end Itaniums, (rx2620s) and we have found them every bit as good performance wise as the Alphas they are replacing.
I recently got back from the VMS Bootcamp in Nashua, one of the sessions compared performance of a particular database under Alpha and Itanium, and showed it to be at least as good on the latter (sorry, can't be more specific, the preso was under NDA). Speak to your friendly neighbourhood HP rep, they should be able to give you some benchark figures.
So if you use the latest OS and latest compilers, IMHO there won't be performance issues.
Have fun,
PJ

[Luk Vandenbussche]

Hi,

Don't forget to check this link

http://www.itanium-integrity.com/

[Alan_152]

I work with most of the HP IA64 linux line in a test lab. They seem to do OK for what I need them to do.

If you have an issue, I'd be happy to try and help.

[Matthew Ghofrani]

Thanks a lot everyone for your inputs, I plan to keep this thread alive to see if I get more inputs.

[Matthew Ghofrani]

I was told about the following link. Thought I should share with all you as well.


Next are some links regarding Itanium-integrity success stories.

http://h71028.www7.hp.com/enterprise/cache/280683-0-0-0-121.html

[Gene Laoyan]

I have to give my 2cents in. We have ordered a new rx8640 and will be in sometime in July. All we kept hearing is about how the second biggest investor in the Itanium market (SGI) is going bellyup. Really at this point, HP is the only "Big" backer of the product. Of course HP said they dumped 10Billion dollars into the market for the Itanium but that still makes me uneasy. Word of mouth is saying it's going the way of the Alpha. I'm not knocking this down or anything but that's what I am hearing and from the stories and reports of SGI, what is one supposed to think at this point. Don't get me wrong, I like the Integrity Servers and If we had the $$$ I would have opted for a Superdome or two.

[Cameron Rooke]

HP decided years ago to use a third party to develop its processors as opposed to making them themselves. They partnered with Intel to develop a new 64-bit processor architecture and came up with Itanium EPIC. It doesn't surprise me that other big players in the server marketplace have so far chosen not to embrace Itanium since they would have their own significant dollars invested in their own chip technologies. HP recognized that RISC can only go so far, and with Itanium, they can take processing power even higher. No one raises an eyebrow when IBM sinks billions into their POWER processors, so what makes HP so different? Itanium (EPIC) is a new technology and anyone who expected it to take off day one is foolish. Itanium will make gains - first with HP, then later with other vendors that need the horsepower for their systems and that do not want to spend the R&D dollars to develop new technology. Down the road, HP can focus it's R&D dollars elsewhere and become even more profitable as their competitors try to catch up.

[Duncan Edmonstone]

If you ask me anyone who thinks that processor technology is the component of a system that adds value and differentiates a system must be living in the past. The fact that HP Integrity servers run Intel Itanium processors is kind of irrelevant, cos the value in the systems is realised by what HP build around the processors in terms of chipsets and software etc. Where's the value in having a processor which is very fast at something like the tpmC benchmark, but in a system that never gets utilised more than 30% ? The key to an effective server is how much value you can drive out of it - I don't like the term, but 'sweating the asset' is often used to deescribe this - making sure that all those expensive processors actually get used. HP achieve this through clever virtualisation and partitioning technologies that let you drive the systems up to 70-80% utilisation numbers.

The Internet tech community still gets so hung up on raw performance figures because its easy to measure compared to other meausrements of value - its like owning a car - 'my sports car has a top speed of 150mph, your people carrier is just 110mph' - so what? You can only get 2 people in yours, I can get 6 in mine, or I can convert the back seats to flat so I can transport goods as well (flexibility drives value!). Luckily the chattering masses on the various web sites ( like theregister or zdnet etc.), don't actually get to make IT decisions in general - they just spend all their time talking about it... Most intelligent CIOs can make a business decision based on more than anm acquisition costs and some speeds & feeds, and as such I'm sure that HP Integrity servers will have a bright future.

HTH

Duncan

(who actually does drive a 2 seater roadster , not a people carrier!)

[Matthew Ghofrani]

Thanks all for your inputs.

I still like to hear more opinions as we have not make a decision yet.

Matthew From Boston

[Robert Gezelter · Accepted solution]

Matthew,

I have been following the IA-64/Itanium issue for many years (I even remember being rebuffed by a representitive when I asked for details about the instruction set: the response was that it was the greatest thing since [canned beer, sliced bread] but tha the instruction set was confidential).

Things have come a long way since then. The situation with the IA-64 has followed a curve that we have all seen before, from any number of CPU architecture vendors.

When a new architecture is announced, and first shipped, the cup in the raw speed/capacity/capability race remains with the older architecture, it has the advantage of implementation experience.

In my own personal experience, I saw this during Digital's sequential release of the PDP-11/VAX, and VAX/ALPHA. When then Compaq announced the decision to adopt IA-64 as its next architecture, I did a quick re-review of IA-64 (the details had in the interim been released), and published an item on the www that I expected the migration of OpenVMS to be a significant effort, but not a major problem. This article was in response to a number of messages that had appeared in comp.os.vms. My concern was not technical issues, it was business issues. This article is still available at http://www.rlgsc.com/alphaitanium.html

I also presented a session at the 2001 Compaq
Enterprise Technology Symposium on the issues involved in migrating applications from Alpha to Itanium from an OpenVMS perspective, "The Third Porting: Applying Past Lessons to the Alpha/Itanium Transition". In this session, I identified the similarities between Alpha and IA-64 on a data level, and projected that the factors that had caused the most trouble in the change from VAX to Alpha were simply not present in a transition from Alpha to Integrity. The session was written from the perspective of an OpenVMS user, but most of the comments are applicable to other operating systems, including Tru64 and HPUX. The notes from this session are available at http://www.rlgsc.com/cets/2001/1620.html

When HP transferred it's stake in the actual chips to Intel, and reallocated its resources to server and system development, it resolved a qualm I had had about the adoption of IA-64 by Compaq, when HP had a privileged position, and that that could become a business issue. I published comments on that issue on OSNews.com, at http://www.osnews.com/story.php?news_id=9191

In short, early adoptors of any technology (read the history of the introduction of jet powered aircraft) almost invariably face the instant catcalls of users who decide that the optimized existing technology is a better bet than the unoptimized new technology.

It is only later, when the new technology begins to assimilate the improvements inevitable in each successive generation, that the benefits of the new approach become apparent (how many large piston airliners have you flown in recently).

My advice to clients in 2001 and now remains the same, "Don't speculate, do science. Buy a samll, workstation/departmental server system (rx-class in the case of Integrity), and develop an experience base with the technlogy and its use. Then we can talk about larger systems (In the OpenVMS world, there is a great deal of latitude on whether the best solution is a collection of relatively small boxes in an OpenVMS cluster, or a single Superdome, or something in between the two).

My apologies if this post is a little long-winded.

- Bob Gezelter, http://www.rlgsc.com

[Matthew Ghofrani]

Robert;

Thanks so much for your inputs & sharing your personal experience. Reading it was very informative.