Oracle stop development on itanium

[Ismail Azad]

Hi,

Well "open source" and the "race to the bottom"... Interesting point... Since the socio economic concept from where the term is derived minorly relates to cost... People would now want to "race to the middle"...

Regards
Ismail Azad

[AQadir_1]

I think oracle has advantage, changing hardware is more easier than changing db and application... just do exp and imp and you are on new hardware. :) ... but i dont like this move from oracle. Because as per my experience customers are very happy with oracle on hp-ux rather than going for x86 platforms

[IS.Danil]

I am always wondering what is OpenVMS market place, for which application it is used now, who will buy OpenVMS?
And now, I am wondering what will be the
HP-UX market place? How will it be changed? As i know Oracle application, especially OracleDB products are the main target for HP-UX. What applications will be the target forhp-UX? So many questions...

[Alzhy]

I would call the ongoing RACE - a race to be "Smart". Smart because as Nicolas Carr is being proved "IT does not matter" anymore. IT is becoming commoditized and those who are SMART -- wins.

FUDs, POV's abound and will continue to abound around the Linux environment. Pundits, self-annointed experts, technical "shysters" if allowed to dominate your trains of thought will consume you good.

Any IT system - whether Open Source based (Linux) or the proprietary ones WILL always have pitfalls and obstacles ahead. How does one mitigate the risks?

- test, test, test

[Steven E. Protter]

Amazing how you can talk for 45 minutes to a reporter and they get very little right.

What I actually said is if HP wants to call Oracle's Itanium bluff, they could port the OS to x86.

I don't see that as a productive move. Coming out with an alternative hardware/software stack that costs less due to use of open source database technology would be a great way to go.

Anyway, there are other databases that work on HP-UX. I hope HP provides an end to end stack that competes and lets customers save money. They'll also need to get more aggressive on Itanium pricing.

SEP

[Duncan Edmonstone]

So referring back to my post dated Mar 25, 2011 16:46:52, I said:

> You see if Oracle finds this business practice works, you can bet they do it again... I doubt they would actually drop support for Oracle on Linux/x86, but I can forsee something like:
>
> i) Increase Oracle core factor for x86 cores from 0.5 to 1.0
>
> ii) If you want to run Oracle on anything other than Solaris/SPARC or OEL/x86 on Oracle Tin, then your support costs are 2x because "we don't control the stack"

It would seem that my prophecy there is already showing signs of coming true. For those with Oracle Support/Metalink access running Oracle on Linux I suggest you read article 1089399.1. I can't post the content here, but the gist of it is, no support for ASMlib on the RHEL6 kernel... you must run the Oracle Linux kernel...

A sign of things to come??

Again, my opinion only - not those of my employer...

HTH

Duncan

[Alzhy]

Duncan,

ASMlib -- yes - it seems no more updates for RHEL 6 and derivatives.

But what is ASMLib anyway? What is its value?

It is simply just a "nice" added layer between ASM and the Storage Stack of linux -- so multipathed disk devices can be pre-allocated, ownerships and headers (block ids) pre-set to be "ASM-ready". It also makes management easy as regardless of how the disk devices come up named -- it will always have a "name" in the eyes of ASM.

My view (and perhaps a growing number as well) is - it is really not needed any more.

Why?

There's UDEV rules where one can name disks friendlier and set ownerships.

But unbeknownst to most, device mapper multipath now offers the option to custom name multipathed disks as well as set ownership as well as access mode -- negating the need for ASMLib and UDEV so the dik devices can be straight used under ASM via "ASM_DIsksring" in ASM.

For example, in /etc/multipath.conf - a Linux Admin can merely set :

multipath {
wwid 3600508b40006a5d50000f00001160000
alias SAP_ASM_disk01
mode 0644
uid 501
gid 501
}


after presenting the SAN disk and ging through online disk presentation commands (multipath -v2, multipath -ll) -- there will be a disk entry named:

/dev/mapper/SAP_ASM_disk01

and ready to be partitioned (always a best practice to parition wholly) and utilized in ASM (via ASM_DISKSTING "/dev/mapper/SAP_ASM_disk*") as

/dev/mapper/SAP_ASM_disk01p1

So there you go -- no need for ASMLib. Unless there are other "performance/efficiency related rationale that Oracle will disclose" for ASMLib that others seem to unsuccessfully prove.

Note the above scheme augurs well for RAC and cluster environments as well -- easier naming of Oracle disks and less complex I may add.


Cheers!

[Alzhy]

In simplistic terms:

ASMlib allows for easy identification of disks used by Oracle ASM.

Simply doing a "blkid" will identify all disks that are marked as ASM used.

Without ASMlib, you'll have no INSTANT way of finding out which disks are under Oracle ASM control.

The above suggested standard however can be in lieue OF ASMLib -- which again people are finding out is NOT REALLY a key ingredient for a successful Oracle on Linux deployment.

[Duncan Edmonstone]

Alzhy,

I think you missed my point - I'm not particularly interested in the value of ASMlib or otherwise, it was there to illustrate that Oracle are already starting to differentiate what they offer on Oracle Linux vs. RedHat or SUSE Linux.

How long before this extends to other components? Initially I could see it being other kernel modules, like those in the RAC hangcheck timer or the ACFS kernel modules
(RedHat have even handed Oracle an 'excuse' to do this with their code obfuscation).

This is one case where TBH I'd be very happy to be proven wrong, but right now I don't see how "OracleDB on non-Oracle Linux/x86" is significantly "safe" as a long term platform from Larry Ellison's standard wallet gouging tactics...

Again - my opinions only.

Duncan

[rariasn]

Hi:

HP Superdome customers are facing costly "forklift upgrades" when upgrading from dead-end PA-RISC and Itanium processors and HP-UX.


http://www.oracle.com/us/corporate/features/cash-for-clunkers-295403.html

rgs,

[Wim Rombauts]

Rariasn,

Have you done any effort to effectively compare the price (and performance) of these Solaris systems with the latest HP-UX systems ?
This type of talk is allways done between vendors. It's always a huge performance benifit if you upgrade a 5 year old system to whatever new hardware, for a fraction of the support price of the old system.

[Prokopets]

Rariasn: two years ago i worked in one company that was an Oracle's partner. So, i can say that discounts on M-series servers were at least 65-70%. As i can see - Larry is _very_ greedy.

Regards,
Philipp.

[Alzhy]

Dunc..

I don't think Larry O will not be that shrewed to lock out all Non Oracle Linux environments our of future versions of the Oracle mega stack specially its RDBMS. They make a LOT of money on RHEL Oracle deployments but I hear ya good sir. Besides - even if Oracle Linux is now an independent "fork" of RHEL having jumped the gun on RHEL 6.0 --OEL (which is still a downstream of RHEL 5) is still officially supported. RHEL 6.X is I thought a gentle reaction to Oracle's agressive fork and the kernel version there is now just a tad below Linus' releases.

Soon I think RHEL 6.X will be current for THOSE adventurers who wish to dabble in the latest and greatet Linux kernel tweaks in 2.6.38 -- which includes the Google TCP/IP Stack, VFS and the supposedly DB accelerating process agrregating processes, BATMAN and "transparent huge pages" -- deep changes that will wring the last ounces of performance of the upcoming 16-core Bulldozers from AMD and the Sandy Bridge Xeons from INTEL.

The explanation above for ASMLib's demise in RHEL 6 is to appease FUDs about those ASMlib dependent shops - re: the picture with ASM sir.

Cheers everyone.. not all is lost for Itanium though -- there's still Ingress/Informix, DB2, Sybase, MySQL, PostGress/EnterpriseDB, SAP, SAS, MicroFocus -- so unless they join the Itanium-away bandwagon -- better think about your next technology refresh.

[tungnt]

I'm using HP PA-RISC with Oracle DB. This time i'm preparing move my system to Itanium.
But now, i have to change my mind. I'm wavering between using Sun, IBM with oracle and using x86 with Oracle RAC.
So pls give me some advices.

[Andrey Chipalyuk]

Alzhy,

My experience show me that Linux is still difficult for using and have not enough quality documentations.
For example, typically business task. I want dynamically to add the disk drive into system.

HP-UX:

# ioscan -fnNC disk

And it is all!!!

Linux:

If we use the HP ProLiant servers:

# hp_rescan -a

For other vendors, for example (sometimes donâ t works):

# echo "- - -" >/sys/class/scsi_host/host{N}/scan

Or

# ql-dynamic-tgt-lun-disc.sh (Script file used to scan the QLogic)

Or

# rescan-scsi-bus.sh

Then I must change settings of multipath-tools (multipath.conf).
For Qlogic driver I must disable failover settings.

ql2xfailover=0

I must change some other settings for Qlogic driver or Emulex.
I must do:

# multipath

And I must verify:

# multipath -ll

And it not all!!!
Russian slang call this is haemorrhoids.

[Wim Rombauts]

tungnt, have you also considered DB2 or EnterprisDB ?

Why we are really reluctant to run Oracle on other hardware : We now use Secure Resource Partitions, which (until Oracle also changes that) allows us to have to run and monitor only 1 hardware, while only paying licenses for the size of the Resource Partition. (Don't begin asking why Oracle allows this technogy, while VMWare and RedHat virtualization is NOT allowed ...).
So, I now only need 2 Enterprise edition licenses on a 2-ndde cluster with 1-socket quad cores.
Switching to AIX will be an expensive move, and a complete new technology for us, but Oracle licensing will be about the same.
Swithcing to non-Oracle linux (which we already have, including clustering and vitrualization) leaves us with a least quad core processors without any allowed option to limit the database CPU resources, so we will end up with having to buy 2 additional Entrprise licenses, inclding 2 additional RAC licenses -> No thanks.
Switching to Oracle linux will bring us new clustering en virtualization technologies to learn jst to be able to limiot the licensing to what we need and now have. And it will also put us completely in the hans of Oracle -> No thanks.

So if Oracle stops with software on HP-UX, there is no option for us than to stop with Oracle.

[Ismail Azad]

Hi,

And yes I see this all over the net as someone pointed out....

"This will change in early Q2 when EnterpriseDB introduces full support for the HP-UX platform via our Postgres Plus Advanced Server."

So it looks like they are averse to the Itanium "away bandwagon".

"If a window of opportunity appears, don't pull down the shade".

Regards
Ismail Azad

[Alzhy]

@Andrey C wrote:

"My experience show me that Linux is still difficult for using and have not enough quality documentations. For example, typically business task. I want dynamically to add the disk drive into system. "

What's difficult with scanning the FC/SCSI Bus? After the scan - you will have your multipath device pronto -- no additional step...

The kernel/multipath configuration you only do once on first presentation (Think if it as configuraring multipath settings -- you do not need to do this everytime you present new LUNs sir!).. The other 2 commands -- multipath and multipath -ll -- are just VERIFICATORY commands sir just like listing all your disks in other UNICes... and custom naming disk devices? see my illustration several posts up sir...

Proost!
Spachiva!

[Andrey Chipalyuk]

Alzhy,

To understand and learn this procedure for HP-UX you should read only the one manual about ioscan. It is easy.
When you want to do similar for the Linux you should read the some manual from different vendors.
For the Linux you need to read about multipath-tools, Qlogic or Emulex driver and about scanning devices. There is not one quality manual. You must search information in the different sources.
It is necessary more time and efforts.

Regards,
Andrey

[Alzhy]

Linux I/O Stack including multipathing (native), storage management (volume manager and filesystems) -- are CONSIDERABLY easier and simpler comrade Sergey.

HP-UX for instance (11.31). ioscan and scsimgr and fcmsutil(fcdutil) - how does it compare to the very easy to learn and be familiar with multipathd and the scsi stack on Linux?

Ditto with the uber-complex Solaris stack -- devfsadm, cfgadm, luxadm...


But to each his own I guess.

[Tony Bosco]

As Wim Rombauts said, many of us have written to the Oracle address highlighting the unreliable direction of Oracle. Some months back, to promote OEL, it was mentioned that RHEL is four years back and OEL is fast, then comes the Itanium processor license form factor increased from 0.5 to 1, which to me is an indication that Oracle consider Itanium processors are powerful now and now withdrawal of development, when both Intel and HP have shown their strategy for the next two generations, Paulson and Kittson and both contradicting Oracle claims.

It looks like Oracle is more aware about Intel & HP strategy and it is like other vendors saying that Oracle will not develop next version of Database and stopping support on their own, when Oracle themselves are having a roadmap.

It is funny to see Oracle strategy is revolving around Redhat and Microsoft withdrawing Itanium support and using these as the reason for their decisions.

As Wim said, I will opt out of Oracle due to this unreliable and arrogant attitude rather than moving out of Itanium Platform.

All customers are taken for a ride by Oracle and not sure what is next in store for Oracle customers.

[Viktor Balogh]

here is a survey i recently found:

http://survey.gabrielconsultinggroup.com/limesurvey/index.php?sid=73634〈=en

[Joe Quijada]

We already have the Itanium Blade Servers running hpux-11.31 in place and installed Oracle Database 10g. Our tests indicates that processes are extremely much faster than our PA-RISC processors running Oracle Database 10g. Here is our solution, instead of changing our hardware platform, we are pushing into changing the Database to DB2 instead, and get out of selfish arrogant attitude of Oracle. Hope everyone out there wakes up.

[Shibin_2]

Recent comment about Itanium from Intel's Senior VP and General Manager Thomas M Kilroy.

The Xeon E7's performance surpassing Intel's Itanium processor , alsoaimed at high end servers, has raised questions about whether it will replace the Itanium.

"The E7 platform's stepped ahead in some cases but Itanium has its own ecosystem. There are mission-critical usages of the Itanium all across the globe. So if you look at stock exchanges, they may be running on Itanium. We are very much committed to Itanium. Even though the volumes are small, it is steady and a very important customer base. We can't confuse strategic importance to volume," says Kilroy.

[Turgay Cavdar]

Here is the HP-UX 11i v3 Update 8' releas notes : http://h20000.www2.hp.com/bc/docs/support/SupportManual/c02751569/c02751569.pdf

Some products related to oracle removed from the OEs.

â ¢ HP Serviceguard Extension for Oracle E-Business Suite (T2260AA): For the March 2011 release, this product has been removed from the OEs. In September 2010, it was added to the OEs in error. See HP Serviceguard Extension for Oracle E-business Suite B.01.00 Release Notes at http://hp.com/go/hpux-serviceguard-docs.
â ¢ HP Serviceguard Toolkit for Oracle Data Guard (T2259AA): For the March 2011 release, this product has been removed from the OEs. In September 2010, it was added to the OEs in error. See HP Serviceguard Toolkit for Oracle Data Guard B.01.00 Release Notes at http://
hp.com/go/hpux-serviceguard-docs.