Linux memory optimisation in 64bit IBM processor

Hi All,
Its my first thread in my SDN.
We are about to migrate a new system in our environment
The new hardware will be IBM x3500 64bit with 20GB and for some unavoidable reasons we are planning to the use this system as 32 bit OS
I am aware if we use the system as 32 bit we have to compromise the memory and performance. But I like to know will this cause a performance problem in SAP or we will not be able to avail the complete usage of memory in the applications and also in case of hardware issue will IBM support us?
Kindly give your thoughts and with your inputs i can take it up further.
Regards

HariniSAP wrote:>
> Hi All,
>
> Its my first thread in my SDN.
>
> We are about to migrate a new system in our environment
>
> The new hardware will be IBM x3500 64bit with 20GB and for some unavoidable reasons we are planning to the use this system as 32 bit OS
>
> I am aware if we use the system as 32 bit we have to compromise the memory and performance. But I like to know will this cause a performance problem in SAP or we will not be able to avail the complete usage of memory in the applications and also in case of hardware issue will IBM support us?
Hi there,
since 32-Bit support is basically just not there for most current SAP product releases, you won't even get to the point where performance would be your concern.
A new server installation, with current product versions - there's no way to do this right on 32-Bit.
regards,
Lars

Similar Messages

Is Linux X86-64(AMD) 64bit OS certified with 10.1.2.0.2?

Hi,
Is Linux X86-64(AMD) 64bit OS certified with 10.1.2.0.2 ?
I found from the Metalink->Certify that it is certified and that 10.1.2.0.2 - 32 bit can be used with Linux X86-64(AMD) 64bit OS.
But I found the following too in the Metalink->Certify
Processor Chip Information:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Information from Certify Matrix:
Oracle Linux x86-64 support on AMD64/EM64T and Intel Processor Chips will only
be on chipsets that adhere to the x86-64 architecture with supported Linux
releases. Running Oracle Database 32-bit on AMD64/EM64T (64-bit OS) is not
supported.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Does this mean that Oracle Databse 32-bit on AMD64/EM64T (64-bit OS) is not supported or the Application Server 10.1.2.0.2 is not supported??
Thanks in advance.
Regards,
Sindhiya V.

Small addendum:
There is a separate Oracle RDBMS release for the Linux AMD64/EMT64 platform which is what you found in the certification matrix. The standard 32-bit Oracle code shouldn't be used on it (nor would you really want to regardless).
As for the app server, it is the same distribution for both the standard 32-bit x86 and the AMD64/EMT64 platforms. A majority of it will run in 32-bit mode (linux32 bash is your friend); the salient difference which may or may not make a difference in your infrastructure: it supports >4GB JVM's with a 64-bit JDK.
Geoff

Memory optimisation

Hi,
My question is the following : What type of memory optimisation does TimesTen do ? If this is used as cache for an oracle database- is the memory footprint really comparable to a relational database - or is there some type of internal optimisation i.e. by indexing or reformatting the internal representation. For example, if I want to cache a 500GB database and I only have 50 GB RAM - does that really mean I can only have 10% of my data in TimesTen ?
For example, if I have one table with 1MM rows and one of the columns only has only two distinct values - will TimesTen could do optimisation and only store those distinct values once - thus reducing the required foortprint.
Dirk

No, TimesTen does not perform the type of optimisations that you are describing. It really is an in-memory relational database (and can be used as a standalone database without Oracle in fact). The common experience is that data tends to take up more space in TimesTen due to the different storage organisation (optimised for performance); a table that occupies 1 GB in Oracle may easily take 2 GB in TimesTen (but this is very schema dependant). Conversely, indexes tend to be quite a lot smaller in TimesTen than they are in Oracle so there is some degree of swings and roundabouts.
Chris

Can I upgrade my 2006 iMac4,1 with Intel Core Duo with 1.5GB memory, 667MHz Bus Speed, 2GHz processor from 10.4.11 to the newer OS systems? I've been told my machine is now "vintage" or obsolete...

Can I upgrade my 2006 iMac 4,1 with Intel Core Duo w/1/5GB memory, 667GHz Bus Speed, 2GHz Processor Speed to the newer OS systems?
I have been told my machine is now "vintage or obsolete" and can't be upgraded to the newest OS software.

The maximum OS for that iMac is 10.6.8.
(64450)

All 4 Planning applications constrained to 4GB Linux memory ?

Hi,
We run Planning on Windows 2003 server and Essbase on Linux.
We have 2 planning applications, Hyperion's Workforce and CapEx applications - making a total of four applications.
Is it true that these four applications all have to share 3 GB of memory (the limiited memory to one Essbase application) ?
Do we use EAS to manually control how much cache and memory to allocate to each of these applications?

Hi,
I take it you are talking about the essbase databases, they all are under one essbase application per planning application this is how planning has been designed, so yes they will share the same maximum for one essbase application.
You could look at moving to 64bit essbase this will give you the option of increasing the available memory to the essbase application process.
Cheers
John
http://john-goodwin.blogspot.com/

Oracle 8.1.5 Linux Memory Leak?

We have been running Oracle 8.1.5 Server and Client on Redhat 6.1
kernel 2.2.12-20
glibc-2.1.2-11
Blackdown's JDK 1.2 RC3
We are having bad memory leaks and we have verified that they are
not in our source code using a memory profiler for Java-linux.
We are using the JDBC calls and the OCI JDBC driver.
We think that the memory leak is in libocijdbc8.so but we're not
sure. It could also be in glibc but the 2.1.2-11 version should
be pretty stable. It could also be in the Blackdown jvm.
Anyone have any ideas?
null

Oracle8iR2's JDBC works fine with Oracle8.1.5.
You can download the new JDBC driver.
null

ODBC Linux Memory Problem

Hi,
I seem to have a big problem with an application I've made.
The application connects to a Oracle Database, fills the database with arround 1GB of information. Once the database is filled my application launches a series of queries to test performance.
My problem is when working with Oracle 11 in Linux (Fedora and Oracle Enterprise Linux). Apparently I can't seem to get past the database filling process. The free memory vanishes quite fast, eaten up between my application proces' and Oracle's. Now, I know the problem is not with my application because the same code works perfectly in the following databases:
Windows:
- Oracle 10R2
- Oracle 11g
- PostgreSQL 8.3
- MySQL 5.1
- MySQL 6-alpha
- SQL Server 2005 Express Edition
Linux:
- PostgreSQL 8.3
- MySQL 5.1
- MySQL 6-alpha
My application actually connects to the database and starts executing the inserts successfully. It's just that at some point the memory gets filled up, pretty quickly actually. I've watched the through vmstat. The first one to fall is the free memory, then follows the cache memory. And then it starts filling up swap. When it reaches to 2GB of used swap the behavior differs with the Linux Distribution: Fedora Core 8 crashes horribly, in Oracle Enterprise Linux at some point my application's process goes to "sleep" and I seem to still be able to do other things.
To connect to the database I use the [OTLv4 library|otl.sourceforge.net]. One might think that my problem is with OTL, but I use the same OTL in all those cases, and unless it's an issue between Oracle in Linux and OTL I don't think it's an OTL issue.
Can anyone push me in some direction ?
Thanks.

Well, you seem to have a memory leak in your application, assuming nothing else is running. Check out the top(1) program; you can get it to sort the displayed processes according to how much %MEM (percent of memory) they are using. First entry is the culprit.
The best way I've found to locate a memory leak is to run your program using valgrind(1). This tool runs your application in a virtual CPU built in to valgrind(1). It can detect tons of different memory usage issues, from using an un-initialized variable to leaked memory.
Of course, you may not have a memory problem. Does you program generate huge bursts of I/O or is the I/O more evenly paced? Kernel I/O backlog can consume memory quickly. While your application is running, open two more windows. In one, run the command:
$ watch -n3 -d cat /proc/meminfo
and in the other:
$ watch -n3 -d cat /proc/slabinfo
Try to figure out what memory usage is steadily increasing; that can tell us if the app is leaking or the I/O channels just can't keep up.
Hope this gives you some ideas about how to proceed.

Linux memory ballooning problem

Hi all, my customer has a MS HyperV 2012 (not R2) cluster for virtualization purposes. Stumbled in dynamic memory allocation I just had a try, reading around about supported configurations etc. As per what i understand, for a 2012 environment the only
concern is to set identical startup and maximum ram values.
Once the vm is booted, after a while, the ram amount start to decrease, meanwhyle in terminal some "out of memory kill process" starts and the amount returns into the original max value.
Any idea or experience about that scenario? The RHEL release is 6.4, integration services embedded.
Tanks in advance.

~
~
. . . , the udevs ones is really interesting.
~
~
But use rules by
Nikolay Pushkarev :
RHEL
6.5 / CentOS 6.5 ; RHEL 6.4 / CentOS 6.4 include support for Hyper-V drivers
==
Following rule works slightly faster for me (assuming that memory0 bank always in online state):
SUBSYSTEM=="memory", ACTION=="add", DEVPATH=="/devices/system/memory/memory[1-9]*", RUN+="/bin/cp /sys$devpath/../memory0/state /sys$devpath/state"
And one more thing, udev solution doesn't work in 32-bit kernel architecture, only for 64-bit. Is this by design or yet another bug?
==
P.S.
And see topic
Dynamic Memory on Linux VM

Fast PC yet still problems with memory optimisation on CS3

I have a Dell 690 with eight 3GHZ processors and 3 Gig of RAM.
I'm editing a half-hour documentary with lots of video audio and still files.
Often when I create a new title caption, I receive a warning that memory is low. And now it's freezing up all the time when I'm animating stills (particularly when I apply a heavy effect like Gaussian Blur).
Do I need more RAM? Or is Premiere CS3 just as lacking as the previous versions, when it comes to editing big projects?
ed

Two suggestions:
1) make sure that under the Preferences options, there's a setting somewhere that says something to the effect of optimize for Performance vs Memory. (By default, it's set to performance). Switch that to memory.
2) Consider using Vista 64. That will allow any 32 bit process to access up to 4 gigs of virtual address space instead of 3 gigs. This should help considerably... plus, you could throw more memory into the machine which would help if you're running multiple concurrent applications (case in point, if you're attempting to use Dynamic Link between AE & PPro).

Video will not open OS 10.6.8 Safari 5.1.10, youtube page freezes the system. Activity moniy=tor shows all memory being used up, but processor use is very low

Video will not open. OS 10.6.8, safari 5.1.10. problem is very noticeable when opening youtube main page. stalls complete system. tried reloading most recent version of flash. Activity monitor identifies all memory is being used to open page, no processor usage and transfer rate does not look out of line. 4 GB memory. Macbookpro6.2, i5, 2.4 GHz. Individual Video will open in facebook. Some video will not open in Yahoo.

Video will not open. OS 10.6.8, safari 5.1.10. problem is very noticeable when opening youtube main page. stalls complete system. tried reloading most recent version of flash. Activity monitor identifies all memory is being used to open page, no processor usage and transfer rate does not look out of line. 4 GB memory. Macbookpro6.2, i5, 2.4 GHz. Individual Video will open in facebook. Some video will not open in Yahoo.

Linux memory settings

We installed the suite on a linux server in a managed configuration (AdminServer, soa_server1, bam_server1).
At first we didnot change the memory settings so all three process ran with memory settings:
-Xms512m -Xmx1024m -XX:PermSize=128m -XX:MaxPermSize=512m
With these settings the suite ran OK at first but gradually the performance detoriated.
Using jconsole I looked at the memory consumption of the processes and the soa_server1 was using up all memory while the AdminServer and bam_server1 had plenty of memory.
I changed the memory setting specific to the servers:
AdminServer, bam_server1
-Xms768m -Xmx768m -XX:PermSize=128m -XX:MaxPermSize=256m
soa_server1
-Xms1536m -Xmx1536m -XX:PermSize=128m -XX:MaxPermSize=512m
The process have been running stable upto now.
The server:
OS: oracle enterprise server release 4 update 7
Cpu: 4*E7340 @ 2.4 Ghz
Ram: 8 Gb
Any other experiences with memory settings on linux and/or using a single server setup vs the shared configuration?
Gr,
Gert Jan Kersten

About rebuilding kernel
on linux 7.0 :
1. Try to look at file /usr/src/linux/include/asm/shmparam.h
Update parameters SHMMAX, SHMMIN, SHMMNI, SHMSEG.
SHMMAX is set to 50% of RAM
SHMMIN 1
SHMMNI 100
SHMSEG 10
2. edit file /usr/src/linux/include/linux/sem.h
and set param
SEMMSL 250
SEMOPM 100
SEMVMX 32767
3. After then you must configure kernel, and set modules
and drivers when you need.
>cd /usr/src/linux
>make xconfig
4. Save configure.
5. >cd /usr/src/linux
>make clean
>make dep
>make bzImage
6. Command make bzImage create new kernel in directory
/usr/src/linux/arch/i386/boot
there is file bzImage
7. copy new kernel to /boot directory
8. you must edit /etc/lilo.cfg and add new record for
new kernel.
9. write lilo to disk
>cd /etc
> lilo -c
10. Be shure that you can boot old kernel !!!!!
11. restart.
12. After correct reboot new kernel you can see to the
/proc/sys/kernel/shmmax
there is parameter about max shared memory.
on Linux 7.1
is different file path and name for edit.
/usr/src/linux-2.4/include/linux/sem.h
/usr/src/linux-2.4/include/linux/shm.h
Hi.

Why Berkeley DB caches entire ".db" file to Linux memory?

Hi,
We use the sample codes from the Berkeley documents to write a large amount of data to Berkeley DB in Linux. We found following problems:
1. System free memory drops quickly as more (key,data) pairs are written into db.
2. When we delete ".db" file, the memory is reclaimed. The system free memory increases to normal level.
3. When we run the same program in Windows, there is no such problem.
It looks that Linux cache the whole ".db" file to the system memory.
It is basically the same problem from this post
Berkeley DB is taking up too much memory
We have tried to use different flags such as DB_PRIVATE or DB_SYSTEM_MEM. It doesn't seems to work.
Many thanks for your help.

Hi there,
841393 wrote:
We use the sample codes from the Berkeley documents to write a large amount of data to Berkeley DB in Linux. We found following problems:What sample codes are you talking about?
What flags are you using when opening the database and the environment?
Do you want for the database to be stored in memory or on disk?
Bogdan Coman

Free memory apps for 32-bit processor

Dear All
There are some free apps for memory like 'memory clean' in the Mac App Store. However my Macbook has 32-bit processor so I cant use most of them. Are there any free apps for memory for 32-bit processor macbooks?
Thanks
veedeekay

You're not missing anything, since all of those apps are completely useless.

Windows vs Linux memory/process management

hi,
my problem is that my windows process that deals with the Oracle Workflow engine grows so big that the Operative System crashes. Is there a way of getting windows processes to use more RAM memory, or is there a way to avoid it? Maybe with a linux system i could overcome windows memory limits ...
regards,
Ricardo

There are several questions you need to answer before a good response can be provided. At the least
1) Do the SAs maintaining the systems have good OS tuning knowledge;
2) Do the DBAs maintaining the databases have good database/instance-to-OS tuning knowledge;
3) Are the graphic subsystems in Linux on or off;
4) Which file systems are to be used on each environment;
5) Which versions of each OS are being compared?
Publishing benchmarks without prior permission is generally a violation of both Oracle and Micorosoft license agreements. I'm not sure one could obtain permission from Microsoft if the benchmark indicates Windows performance sucks. So any performance benchmarks, other than those published at TPC.org, need to be viewed with suspicion. (And those at tpc.org were built at a significant cost by people who are experts at tuning, at least for specific cases.)
In my informal evaluation, in which I simply installed 10g on a certified version of Linux and of Windows on identical hardware (P2.2 GHz, 2GB RAM, 40GB/5400/2M-cache disk). I did not tune the OS or the application in any way - this is not a benchmark, just an observation relevant to my configuration. I noted the performance was about the same.
However, dropping Linux to runlevel 3 got me about a 40% increase in performance compared to runlevel 5. But ... I did not check on a variety of different hardware configurations.
The answer, my friend, is simple ... try it yourself.

Safari 32bit memory use vs 64bit

I have to switch Safari into 32bit mode for a particular website that uses Adobes FileOpen software for PDFs. I normally have two Safari windows open with about 15 tabs open between them. I know 32bit does not use memory as efficiently as 64bit mode. But, how much difference will I see in memory use? Or is it negligible?

Negligible.
BTW.. if you are running the latest Flash plugin version you shouldn't have to set Safari to open in 32 bit mode anyway... or try troubleshooting by uninstalling your current copy of Flash then reinstalling.
http://kb2.adobe.com/cps/865/cpsid_86551.html#ostype=m,prob1=fnctn,prob2=game,

Linux memory optimisation in 64bit IBM processor

Similar Messages

Maybe you are looking for