Oracle 10g on Windows 2003 x64 Memory Useage
sev...@kcpweb.net
We are running Oracle 10g on a Quad Dual Core Windows 2003 x64
Enterprise Server with 16GB of memory. I notice that the Oracle
process is only using 4.6 GB of Physical RAM and 10GB of Virtual
memory. Performance monitor is showing that the server has 8GB
Physical Memory free. This seems like oracle isn't using the free
memory very efficiently and swapping a lot out to disk, is this
something that can be tuned to boost performance or is it dynamically
performed by the Oracle process?
Thanks for your help, I'm new to Oracle having worked mainly with MS
SQL so go easy on me,
Regards,
Stephen
DA Morgan
What evidence do you have that anything is being swapped to disk?
Go to http://tahiti.oracle.com and search for initialization
parameters that related to memory utilization. It is completely
configurable. But don't change anything based upon only the
information provided above. Bigger does not equal better. You
may have just wasted money putting too much RAM in the server.
--
Daniel A. Morgan
University of Washington
damo...@x.washington.edu (replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org
hpuxrac
If you run oracle on a unix platform of some kind you may have better
luck.
If you have properly installed and configured 64 bit oracle on a 64
bit os this sounds unusual.
joel garry
metalink 657456.992 seems to show your problem exactly. Says "What I
did find though, is whomever setup the Windows server set it up as a
File Server. I removed that option and now my Oracle process is using
6 gigs of memory (5 for SGA, 500 megs for PGA and then some over
head) ...My Page File Usage% is now at 1%."
Whatever that means. What didn't work for the guy was
http://download-west.oracle.com/docs/cd/B14117_01/win.101/b10113/ap_64bit.htm#CHDGFJJD
which tells how to grant the privilege to lock things in memory, so
I'm wondering if the real solution was something else that guy did
while fiddling about.
Since I'm a unix bigot, this is all just academic entertainment for
me.
To find the referenced document in metalink, log in there and click on
Advanced Search, and put 657456.992 in the document id box.
jg
--
@home.com is bogus.
http://www.ossec.net/en/attacking-loganalysis.html#denyhosts
sev...@kcpweb.net
Ill post back my findings.
As for running Oracle on Unix, i dont have the choice, the clients
specified Windows so for now ill just have to work it out.
Stephen
Charles Hooper
Have you enabled large page support on Windows and Oracle, which
causes Windows to use 4MB memory page sizes rather than 4KB memory
pages? If so, I suspect that Performance Monitor is showing incorrect
information. I am running Oracle 10.2.0.2 on Windows 2003 x64, total
page file size for all drives is configured at 3124MB, yet Task
Manager reports that PF usage is 13.3GB. In my case, system
performance is great, so I am not too concerned about the inconsistent
numbers.
Have you verified that the page file is actually 10GB in size?
Charles Hooper
IT Manager/Oracle DBA
K&M Machine-Fabricating, Inc.
Paul Linehan
Charles Hooper wrote:
> Have you enabled large page support on Windows and Oracle, which
> causes Windows to use 4MB memory page sizes rather than 4KB memory
> pages? If so, I suspect that Performance Monitor is showing incorrect
> information. I am running Oracle 10.2.0.2 on Windows 2003 x64, total
> page file size for all drives is configured at 3124MB, yet Task
> Manager reports that PF usage is 13.3GB. In my case, system
> performance is great, so I am not too concerned about the inconsistent
> numbers.
> Have you verified that the page file is actually 10GB in size?
I have read quite a few of your posts and I think
that:
a) you know what you're talking about
and
b) you genuinely try to help people.
I hope that this info can help you and/or others.
For Windows, there is a nifty/cool/rad/awesome(for Yanks
only) - (choose term which hasn't gone out of fashion, or
better yet, the one which has come back into fashion!!)
Go to www.sysinternals.com and go for procexp (process
explorer).exe - it gives more reliable results than
the Windows performance monitor.
Inconsistent numbers should not happen. The site also
has a number of other useful bits and bobs for anyone
concerned with performance.
Paul...
> Charles Hooper
hjr.p...@gmail.com
Apart from that useful bit of advice, be warned that 64-bit Oracle on
64-bit Windows 2003 that runs on AMD processors will experience a
massive performance hit as memory starts paging maniacally once total
memory consumption hits about 12.8GB. It's a known bug, there's a TAR
out on it, it's allegedly fixed in 10.2.0.3 (not yet tested, however),
and it only affects AMD processor platforms.
Might not have any relevance to the OP, but I thought I'd mention it
just in case...
Charles Hooper
wrote:
> Charles Hooper wrote:
> > Have you enabled large page support on Windows and Oracle, which
> > causes Windows to use 4MB memory page sizes rather than 4KB memory
> > pages? If so, I suspect that Performance Monitor is showing incorrect
> > information. I am running Oracle 10.2.0.2 on Windows 2003 x64, total
> > page file size for all drives is configured at 3124MB, yet Task
> > Manager reports that PF usage is 13.3GB. In my case, system
> > performance is great, so I am not too concerned about the inconsistent
> > numbers.
> > Have you verified that the page file is actually 10GB in size?
>
> I have read quite a few of your posts and I think
> that:
>
> a) you know what you're talking about
>
> and
>
> b) you genuinely try to help people.
>
> I hope that this info can help you and/or others.
>
> For Windows, there is a nifty/cool/rad/awesome(for Yanks
> only) - (choose term which hasn't gone out of fashion, or
> better yet, the one which has come back into fashion!!)
>
> explorer).exe - it gives more reliable results than
> the Windows performance monitor.
>
> Inconsistent numbers should not happen. The site also
> has a number of other useful bits and bobs for anyone
> concerned with performance.
>
> Paul...
Thanks for the suggestion - I have not experimented with that utility
in quite a while.
Looking at the page file on the hard drive, it is currently at
3,199,488KB, with an upper limit hard coded at 4,092MB (4,190,208KB).
Process Explorer:
Commit: 13.6GB (roughly 14,260,634KB)
Commit charge current: 14,254,748KB
Commit charge peak: 18,208,828KB
(recorded roughly 60 seconds after the above)
Task Manager:
PF Usage: 13.5GB (roughly 14,155,776KB)
Commit charge total: 14,258,808KB
Commit charge peak: 18,208,828KB
Physical memory total: 16,772,124KB
" " available: 3,798,000KB
Comparing the numbers between Process Explorer (now a Microsoft
product) and Task Manager on Windows 2003 x64 are very consistently
inconsistent.
16,772,124KB
- 3,798,000KB
-------------
12,974,124KB physical memory in use
+14,254,748KB page file in use
-------------
27,228,872KB total memory usage
-or-
12,974,124KB physical memory in use
+ 3,199,488KB actual page file size
-------------
16,173,612KB total memory usage
I *suspect* that the problem has to do with the memory page size being
4096KB for applications using large page support, rather than the
expected 4KB memory page size. There is probably a bit of code in
Windows that does not yet recognize the change (this is a new feature
for Windows), either that or Microsoft should patent the approach of
jamming 14,254,748KB of data into a 3,199,488KB file. :-)
Charles Hooper
Thanks for posting the above. Interesting - I wonder if this issue/
inconsistency is a bit more widespread/different than Oracle
believes. My servers have Intel processors. System performance is
great, so I am not too concerned about the inconsistent numbers. But,
if there is a real problem...
Charles Hooper
> On Aug 9, 6:14 pm, "Paul Linehan" <plinehan__A@T__yahoo__D.OT__COM>
> wrote:
> > I have read quite a few of your posts and I think
> > that:
>
> > a) you know what you're talking about
>
> > and
>
> > b) you genuinely try to help people.
>
> > I hope that this info can help you and/or others.
>
> > only) - (choose term which hasn't gone out of fashion, or
> > better yet, the one which has come back into fashion!!)
>
> > explorer).exe - it gives more reliable results than
>
> > Inconsistent numbers should not happen. The site also
> > has a number of other useful bits and bobs for anyone
> > concerned with performance.
>
> > Paul...
>
> Thanks for the suggestion - I have not experimented with that utility
> in quite a while.
>
> Looking at the page file on the hard drive, it is currently at
> 3,199,488KB, with an upper limit hard coded at 4,092MB (4,190,208KB).
>
> Process Explorer:
> Commit: 13.6GB (roughly 14,260,634KB)
> Commit charge current: 14,254,748KB
> Commit charge peak: 18,208,828KB
>
> (recorded roughly 60 seconds after the above)
> Task Manager:
> PF Usage: 13.5GB (roughly 14,155,776KB)
> Commit charge total: 14,258,808KB
> Commit charge peak: 18,208,828KB
> Physical memory total: 16,772,124KB
> " " available: 3,798,000KB
>
> Comparing the numbers between Process Explorer (now a Microsoft
> inconsistent.
> 16,772,124KB
> - 3,798,000KB
> -------------
> 12,974,124KB physical memory in use
> +14,254,748KB page file in use
> -------------
> 27,228,872KB total memory usage
>
> -or-
>
> 12,974,124KB physical memory in use
> + 3,199,488KB actual page file size
> -------------
> 16,173,612KB total memory usage
>
> I *suspect* that the problem has to do with the memory page size being
> 4096KB for applications using large page support, rather than the
> forWindows), either that or Microsoft should patent the approach of
> jamming 14,254,748KB of data into a 3,199,488KB file. :-)
>
> Charles Hooper
> IT Manager/OracleDBA
> K&M Machine-Fabricating, Inc.
A member of this forum suggested to me a different possible cause for
the inconsistency of the page file usage numbers shown in the Task
Manager and Process Explorer. So, I used Google to search Microsoft's
site for additional clues.
It appears that on Windows XP and Windows 2003 Server that the "PF
Usage" statistic (not present in previous versions of Windows) is NOT
page file usage. Instead, it is total virtual memory usage for
Windows and all applications, which includes both RAM and data stored
in the page file (32 bit applications on 32 bit Windows have access to
either a 2GB or 3GB virtual memory range). The "PF Usage" statistic
is esentially the same as the "Commit Charge: Total" statistic, which
is the size of the virtual memory in use by all processes.
References:
http://www.microsoft.com/technet/prodtechnol/Windows2000Pro/reskit/part6/proch27.mspx?mfr=true
"Commit Charge: Total - Size of virtual memory in use by all
processes, in kilobytes."
http://www.microsoft.com/resources/documentation/windows/xp/all/proddocs/en-us/taskman_monitor_perf_fields_overview.mspx?mfr=true
Commit Charge (K): Memory allocated to programs and the operating
system. Because of memory copied to the paging file, called virtual
memory, the value listed under Peak may exceed the maximum physical
memory. The value for Total is the same as that depicted in the Page
File Usage History graph."
http://www.microsoft.com/resources/documentation/windows/xp/all/proddocs/en-us/taskman_monitor_perf_fields_overview.mspx?mfr=true
"PF Usage: The amount of paging file being used by the system. If your
computer is running near the maximum, you can increase the page file
size."
http://www.microsoft.com/technet/archive/winntas/maintain/monitor/03wntpca.mspx?mfr=true
"Commit Charge Provides information on the total memory used by the
operating system. Total lists all physical and virtual memory
currently in use. Limit lists the total physical and virtual memory
available. Peak lists the maximum memory used by the system since
bootup."
http://www.microsoft.com/technet/prodtechnol/Windows2000Pro/reskit/part6/proch28.mspx?mfr=true
"To see how much virtual memory your Windows 2000 Professional-based
computer uses, start all applications and use Task Manager to see the
Peak Commit Charge value. This value appears in the Commit Charge box
on the Performance tab. Commit charge is the number of pages reserved
for virtual memory that are backed by the paging file. Peak committed
memory is the highest amount of virtual memory (in bytes) that has
been committed over this sample. To be committed, these bytes must
either have a corresponding amount of storage available on disk or in
main memory. Compare this value against the size of the paging file to
determine whether the paging file is sized appropriately."
http://www.microsoft.com/technet/archive/ntwrkstn/reskit/03tools.mspx?mfr=true
Shows "Mem Usage" where later versions of Windows show "PF Usage".
Windows Vista shows "Memory" in the same location. The "Page File"
statistic in Vista, which is currently 1471MB (roughly 1,507,328KB) on
my laptop, matches up very closely with the value 1,512,852KB reported
by "Commit Charge Current" in Process Explorer.
Note that there is a bit of inconsistency in the above quotes, which
are all from various Microsoft sources. From the above, using the
numbers that I previously posted:
14,254,748KB (PF Usage, Commit charge total)
-16,772,124KB (Physical memory total)
+ 3,798,000KB (Physical memory available)
-------------
1,280,624KB (Calculated page file size in use)
This seems to make a bit more sense, as the calculated page file in
use falls below the hard coded maximum page file size of 4,190,208KB.
Thanks to Paul Linehan for suggesting the Process Explorer utility,
Howard Rogers for providing a reference to a related Oracle bug, and
Yong Huang for suggesting that there is possibly a different
explanation for the inconsistency.