AIX

just an idea...

Oracle Architecture and Tuning on AIX (v 2.30).pdf

Code:

Starting with Oracle 11.2.0.2 when AIX 6.1 or AIX 7.1 is detected, Oracle will use O_CIOR option to open a file on JFS2. 
Therefore you should no longer mount the filesystems with mount option –o cio.

When Oracle is open a file in cio-mode your system will not cache this data.
Maybe you need do define a greater SGA to reach a good cache hit:

Code:

SGA_MAX_SIZE
Starting with Oracle 9i, the Oracle SGA size can be dynamically changed. It means the DBA just 
needs to set the maximum amount of memory available to Oracle (SGA_MAX_SIZE) and the initial 
values of the different pools: DB_CACHE_SIZE, SHARED_POOL_SIZE, LARGE_POOL_SIZE etc… 
The size of these individual pools can then be increased or decreased dynamically using the ALTER SYSTEM 
command, provided the total amount of memory used by the pools does not exceed SGA_MAX_SIZE. If LOCK_SGA = TRUE, 
his parameter defines the amount of memory Oracle allocates at DB startup in “one piece”! Also, 
SGA_TARGET is ignored for the purpose of memory allocation in this case.

PS: Our DB-Mashine runs "normal" with 30.000 Page Faults/sec

zero filled page faults is an application behavior (e.g., malloc). Very little a system administrator can do.

File caching is automatic with AIX. Starting with AIX 6.1 the defaults for file caching behavior are pretty to very good.

Code:

              8388608 memory pages
              8058736 lruable pages
              2428935 free pages

24/84 % memory free
80/94 % memory demand pageable (least recently used "able")

Code:

              1404618 pinned pages
               384661 file pages
               384661 client pages
                 66.5 percentage of memory used for computational pages

4/84 % file caching (client == jfs2 files, files = jfs + jfs2, so (nearly) all files are jfs2
14/84 % memory pinned
computational memory is all working memory (application memory, plus program code (text))

Code:

                    5 pending disk I/Os blocked with no pbuf
                    0 paging space I/Os blocked with no psbuf
                 2228 filesystem I/Os blocked with no fsbuf
                 1537 client filesystem I/Os blocked with no fsbuf
                 4155 external pager filesystem I/Os blocked with no fsbuf

Above tells me:
* little raw disk i/o being done and/or sufficient buffers
* no paging space activity
* some jfs activity, and initial fs buffers could be tuned to higher number
* some real nfs/cd/dvd activity needing more buffers
* same for jfs2
In short, file system buffering should be checked.

ioo -a
nfso -a

commands will list the tunable variables.

thanks for all the replays, but i cannot understand 1 things.

why aix tell me there is page faults when all disks are running in asm mode and no one are running with raw/jfs etc...

the vmm take count of disk not formatted with unix filesystem also or take count only with know filesystem.

thanks again
Giulio

"Paging" in AIX stands for any physical i/o. A page fault is simplyy when an application wants to access something in memory that is not there.

A "zero-page" fault is simply the first time a page (generally a 4k concept, but tuning can make frames (the in core page concept) be allocated in 64k, or much larger (16M byte iirc). So, for a zero-page fault, no physical i/o is needed when there is free memory available (as is your case).

Way back when, AIX was the the first UNIX to use "free memory" for file caching, rather than a set number of buffers. This has made dicusssions about paging difficult. On older UNIX (as I do not know if Solaris/HPUX/other *NIX, or Linux use file caching like AIX) the term paging referred to what AIX calls "paging space" i/o.

If I were to visit on-site to look at your situation I would look at, among other things, vmstat -I to see paging space and file paging activity.

Again, from the data above Ido not see a reason to link there is a physical i/o problem. ANd when no physical i/o is needed the paging stats can get quite high.

What would help me, is for your to rephrase your problem - or question. JUst to make sure wee are on the same page, as it were. Statistics alone can be very misleading/confusing.

---------- Post updated at 03:43 PM ---------- Previous update was at 03:34 PM ----------

Code:

                 2228 filesystem I/Os blocked with no fsbuf 
                 1537 client filesystem I/Os blocked with no fsbuf 
                 4155 external pager filesystem I/Os blocked with no fsbuf

My "feeling" regarding these statistics is that the system has some initial start up issues, because the fsbufs are too small initially - but AIX is automatically expanding them.

IIRC the AIX/Oracle document referenced above also talks about how to cause AIX to increase these buffers faster (look for the command ioo in the document).

Note: I could type all the commands here, but that is, imho, a false sense of security - as if all is covered. What I hope is that after reading you will have a new question or questions - and from your new questions I will also better understand your problem. In short, I believe in dialog, not "This is it...", or "wave a magic wand" kind of answers.