I am resolved that the trouble is not enough swap and not enough aio
processes. I think this will significantly improve the system. Spreading
swap will be my next push if I still have trouble. vmtune will be last
which is basically what I wanted to confirm.
AIX vmstat shows
kthr memory page faults cpu
----- ----------- ------------------------ ------------ -----------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
0 0 522715 4737 0 1 1 39 115 0 147 440 151 3 1 92 3
0 2 522715 4646 0 0 0 0 0 0 522 1169 133 3 4 91 2
0 2 522715 4646 0 0 0 0 0 0 472 564 60 0 0 99 0
However topas shows more info about where the pageing is occuring i.e.
good/normal paging as oposed to bad paging as seen below. PgspIn and
PgspOut are the bad kind.
Topas Monitor for host: FOO EVENTS/QUEUES FILE/TTY
Thu Dec 20 18:10:02 2001 Interval: 2 Cswitch 2625 Readch
11505467
Syscall 40623 Writech
883896
Kernel 42.7 |############ | Reads 5506 Rawin
0
User 13.1 |#### | Writes 514 Ttyout
318
Wait 9.3 |### | Forks 179 Igets
9
Idle 34.7 |########## | Execs 177 Namei
6369
Runqueue 7.5 Dirblk
49
Interf KBPS I-Pack O-Pack KB-In KB-Out Waitqueue 2.0
en1 9.4 25.9 16.9 3.1 6.3
en0 1.4 10.9 3.9 0.9 0.5 PAGING MEMORY
Faults 30366 Real,MB
4095
Disk Busy% KBPS TPS KB-Read KB-Writ Steals 0 % Comp
39.2
hdisk0 47.4 574.7 92.3 9.9 564.8 PgspIn 2 % Noncomp
61.4
hdisk1 41.9 572.8 89.8 0.0 572.8 PgspOut 0 % Client
348
hdisk3 1.4 337.3 10.4 0.0 337.3 PageIn 66 2
0992
hdisk10 0.0 0.0 0.0 0.0 0.0 PageOut 229 PAGING
SPACE
hdisk4 0.0 3.9 0.4 0.0 3.9 Sios 118 Size,MB
2992
% Used
22.3
dm_ep_eng(11094) 12.5% PgSp: 1.0mb root % Free
77.6
init (1) 6.0% PgSp: 0.6mb root
ksh (35124) 5.0% PgSp: 0.3mb oracle
ksh (15854) 4.00 10.0.4mb oracle 4 Press "h" for help
screen.
ksh (68298) 4.05 PgSp: 0.3mb oracle 5 Press "q" to quit
program.
sadc (59690) 3.0% PgSp: 0.1mb root
ksh (53722) 3.0% PgSp: 0.3mb oracle
ksh (63038) 2.0% PgSp: 0.3mb oracle
ksh (27478) 2.0% PgSp: 0.3mb oracle
ksh (61336) 2.0% PgSp: 0.3mb oracle
-----Original Message-----
Sent: Thursday, December 20, 2001 3:40 PM
To: Multiple recipients of list ORACLE-L
Ethan,
AIX performs all file i/o via the OS's virtual memory. In other words, the
VM (which is a combination of both the RAM and the Swap) is used for both
executable (OS/user code) and user's memory area as well as for I/O buffers.
This allocation is dynamic and certain upper and lower limits are set by the
parameters in vmtune. At times of high i/o, you will find that the 'fre' mem
(seen in vmstat' goes down drastically and the po/fr/sr goes up drastically
as a result of old memory pages that served file i/o buffers are paged out.
This is normal and in line with what you are observing and is similar to
Solaris.
What you *should* be concerned about is excessive values in 'pi' as this
indicates excessive paging-*in*, probably due to earlier pageouts of live
pages that are still required. You are probably seeing high I/o waits since
there is a lot of writes (paging out) to the swap area. 'iostat' will
probably indicate that disks that host your swap area are almost 100%
loaded. In this case, you should make sure that you (a) spread out swap on
multiple disks, making sure that they are NOT on RAID-5. (b) allocate
dedicated drives to swap areas if possible (no RAID5!).
FWIW, vmstat in Solaris is able to distinguish between pi/po values for
executable, file i/o and 'other' (?) types of paging. Can you find out for
us if AIX also provides this? (Don't have access to an AIX box).
Merry Christmas and a Blessed New Year!
John Kanagaraj
Oracle Applications DBA
DBSoft Inc
(W): 408-970-7002
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Post, Ethan
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
