lrud process slowing down the system

[arvibm]

Hi,

I am facing performance problem with one of our servers.I observed the server goes slow down the moment lrud process gets initiated. lrud process cpu usage in topas output is around 25% and hdisk 0 and hdisk 1(internal disks) gets 100 % utilized. None of the user will be able to connect to server the time lrud process gets initiated. After some time the server starts working perfectly normal and repeats the slowdown behaviour after a interval of around 20-30 minutes.I have changed the following vmtune parameters around one month back.The time lrud process gets initiated i am unable to connect to server from any new session.

#vmtune parameters
/usr/samples/kernel/vmtune -p 5 -P 10 -s 1 -R 32 -f 480 -F 512
/usr/samples/kernel/vmtune -b 200 -B 800

why this lrud process gets initiated on server?why the process slow down the system.


Regards

arvibm

 

[arvibm]

Hi ,


sorry i have posted onw wrong value in vmtune parameters.here is the corect vmtune parametes that i have changed around one month back.

#vmtune parameters
/usr/samples/kernel/vmtune -p 5 -P 10 -t 10 -s 1 -R 8 -f 480 -F 512
/usr/samples/kernel/vmtune -b 200 -B 800

in vmstat output page in and out and wait time increase , the time server goes slow down.

Regards

arvibm

 

[donttreadonme]

lrud is the vmm page stealer - it is awakened when there isn't enough memory and page out the less accessed pages.

 

[arvibm]

Hi,

Thanx for reply.How can i trace out the process consuming maximum amount of memory.The server is p-series 6H1 Server with 4-way processor(600 Mhz each)+ 4 GB Ram.How can i trace a particular oracle process or any other process consuming maximum amount of memory because on oracle all the process are shown against one database user.I am unable to trace out the oracle user who has submitted request from browser.
any suggestions how to resolve the performance problem.Pls let me know if any more output's are required to resolve the problem.
i tried with shutdown database and applications on server but when i checked the shared memory segments using "ipcs -s " it diplays number of shared memory segments associated with application users whereas no process is running with that application user.
Is there any way to clear the shared memory segments depending on Operating system user.
I tried to resolve the problem by increasing paging space from (4+4) to (6+6) but no results.

pls help me out to resolve the problem.

Regards

arvibm

 

[donttreadonme]

Most likely you need to add real physical memory.

You can clear mem segments using the ipcrm -m KEYID.

You might want to look into the vmtune command in /usr/samples/kernel, which could help you with Oracle performance. You also need to look at file cacheing. Oracle caches so you don't want a large amount of OS file cacheing which is just doubling work.

 

[bi]

If you have shut down all the databases and the listener, and you still show shared memory segements used by Oracle with ipcs -s, you could use ipcrm -s <semid> to remove those segements. semid is the number in the second field of the ipcs -s output.

I would also check ipcs -m and use ipcrm -m <id> to remove them. Just make sure the databases are down.

Alternatively, you can boot the box.

 

[arvibm]

Thanks for the reply. I've been watching this server (with topas) all morning and now
I see that the rootvg disks (two local mirored disks) are running at 100% .This problem usually happens in morning between 10:00 a.m. to 12:00 p.m.Server stops responding for a period of around 30-40 minutes.Its very difficult to trace what wrong goes with server in morning hours.i need a help to trace out whats wrong going with server in that particular time duration.Pls let me know if any output is required to analyse the problem.

Regards

arvibm

 

[bi]

I would suspect heavy paging is going on, unless you have some datafiles from the database in rootvg. Did you happen to do an lsps -a during this slowdown? And output for vmstat might also be useful. Are the database files optimally spread across disks?

I would talk to the DBA and/or the application person to find out what's going on in the application during the slow times. Are they running heavy batch programs at this time? Are they doing a database dump at this time? There may be some database tuning that would prevent the slowdown.

And finally, is this related to your problem in thread52-666448?

 

[arvibm]

Hi bi,

The DBA has reduced database buffer size yesterday and i have not faced the slowdown problem today morning.the entire database files are on external disk array(FAStT 500)which is on RAID-5.The DBA is unable to trace out any oracle process responsible for slowdown behaviour.I am unable to issue any command on server at the time of problem.

After changing database buffer size i am not facing problem of slowdown but still in sar output the system seems to be a cpu bound(%usr+%sys >80%).here is the output after changing database buffer size.

# sar -u 2 10

AIX hiltrn 1 5 000893DA4C00 11/15/03

12:01:42 %usr %sys %wio %idle
12:01:44 85 14 0 0
12:01:46 84 14 1 0
12:01:48 83 16 1 0
12:01:50 85 15 0 0
12:01:52 87 13 0 0
12:01:54 85 14 1 0
12:01:57 85 15 0 0
12:01:59 89 10 0 0
12:02:01 87 12 1 0
12:02:03 88 11 1 0

Average 86 13 1 0
#
#
# iostat 2 5

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.1 145.1 20.1 6.9 65.9 7.2

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 2.4 23.6 2.7 10234869 48826107
hdisk0 2.5 27.4 3.3 19662529 48844135
cd0 0.0 0.0 0.0 0 0
dac0 0.0 1193.7 265.7 1917659231 1063629990
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 21.4 1193.7 265.7 1917659231 1063629990

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.0 399.5 79.1 15.7 4.1 1.1

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 0.0 0.0 0.0 0 0
hdisk0 1.0 4.0 1.0 8 0
cd0 0.0 0.0 0.0 0 0
dac0 0.0 2104.8 216.6 1420 2816
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 16.4 2104.8 216.6 1420 2816

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.0 398.0 86.9 11.4 1.6 0.1

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 1.0 16.0 1.0 0 32
hdisk0 1.0 16.0 1.0 0 32
cd0 0.0 0.0 0.0 0 0
dac0 0.0 1646.0 69.0 184 3108
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 14.5 1646.0 69.0 184 3108

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.0 398.0 83.6 12.6 3.0 0.8

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 0.0 2.0 0.5 4 0
hdisk0 0.5 14.0 2.5 28 0
cd0 0.0 0.0 0.0 0 0
dac0 0.0 1864.0 123.0 560 3168
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 24.0 1864.0 123.0 560 3168

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.0 397.5 83.0 12.0 4.4 0.6

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 1.0 6.0 1.5 0 12
hdisk0 1.5 6.0 1.5 0 12
cd0 0.0 0.0 0.0 0 0
dac0 0.0 1310.4 45.9 24 2600
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 5.5 1310.4 45.9 24 2600
#
#
# vmstat 2 10
kthr memory page faults cpu
----- ----------- ------------------------ ------------ -----------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
2 5 822808 2084 0 1 1 646 538 0 790 8425 4783 20 7 66 7
6 0 819347 5119 0 0 0 0 0 0 656 27119 9835 90 10 0 0
6 2 819387 1968 0 0 0 0 0 0 1703 25107 13590 87 13 0 0
8 6 819536 580 0 0 0 262 2089 0 1231 26301 11964 88 12 0 0
7 0 815885 4088 0 0 0 0 0 0 540 25230 8285 91 9 0 0
6 0 816104 3709 0 0 0 0 0 0 553 26087 9487 91 9 0 0
6 0 816548 3178 0 0 0 0 0 0 504 23158 9682 90 10 0 0
10 0 816926 2596 0 0 0 0 0 0 525 21575 9108 91 9 0 0
7 0 819270 512 0 2 0 278 2013 0 564 33864 9465 86 14 0 0
9 1 816325 3685 0 0 0 151 1046 0 520 24261 9527 89 11 0 0
#
#

bi i need ur expert tips on my output.

thanx and regards

arvibm

 

[bi]

Some questions: What kind of server is this? Which OS version? How many CPUs and their speed? How much memory? Which version of Oracle?

From the output of the iostat, am I right in guessing that the entire database is on hdisk5?

 

[bi]

arvibm, Sorry I hadn't noticed that you had already given the server type, CPUs, and memory.

You might want to read through this thread, which has a lot of discussion about performance tuning: thread52-649428.

Another thread you might look at it is thread52-208384 but this is mainly for a much older system than your 6H1.

I would still be interested in knowing if the entire database is on hdisk5. I have seen this CPU bound type behavior when the files for a database are poorly placed. How large is the database?

Also, I would suggest that you ask your DBA to run Oracle's utlbstat/utlestat (or statspack if 9i. utlbstat/utlestat also is available on 9i) during the time of the slowdown. This is an Oracle tool that measures what's going on in a database during a particular time. You start it with utlbstat (a couple of minutes before the slowdown usually starts), let it run during the time when you have the slowdown, then run utlestat, which ends the statistics gathering and writes out a report that the DBA can look at to see what's been going on in the database during that time.

There are also some v$XXX Oracle views that might be helpful, but they show stats from when the instance was started and so you might not really see what the problem is.

 

[arvibm]

Hi Bi,

I have passed on ur suggestions to Oracle DBA.The problem is still not resloved even after changing database buffer size only the timings of problem are changed the problem usually comes in evening hours now.The Server goes slow down for a period of 40 minutes.I have shutdown the two database which are not in use and the server started wqorking perfectly normal.4 oracle instances are running on server and database version is 9.0.2.Yesterday i managed to issue some commands at the slowdown behaviour of server.
#vmstat 2
kthr memory page faults cpu
----- ----------- ------------------------ ------------ -----------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
2 5 1047522 2449 0 1 1 632 847 0 778 14085 4596 22 7 65 7
12 19 1049495 359 0 155 126 790 1269132 2 1458 32909 5061 31 34 0 36
1 29 1050247 0 0 7 7 245 2273165 2 655 4040 3244 1 26 0 73
1 35 1050249 0 0 0 0 1 2318739 2 593 3659 2837 1 25 0 74
1 36 1050290 0 0 9 2 257 2306838 3 613 3681 3254 1 25 0 73
1 34 1050301 0 0 0 0 30 2316664 2 598 3695 2915 1 26 0 74
2 31 1050481 0 0 4 91 20 2294995 2 624 3816 3060 1 25 0 74
1 30 1050494 0 0 0 0 91 2315174 3 597 3753 2975 2 25 0 74
1 30 1050761 0 0 5 144 23 2308270 2 643 3928 3122 1 26 0 73
1 31 1050854 0 0 3 47 154 2309563 2 615 3911 3097 0 26 0 73
1 30 1050864 0 0 0 0 45 2313642 3 629 3746 3045 1 25 0 74
1 32 1051035 0 0 2 92 25 2307364 2 621 4011 3257 1 26 0 73
2 33 1051272 0 0 3 87 131 2303882 2 630 3929 3343 1 26 0 74
1 31 1051418 0 0 2 69 109 2310343 3 620 3904 3163 1 26 0 74
1 32 1051445 0 0 0 0 16 2315887 2 613 3873 3022 1 26 0 74
1 35 1051616 0 0 1 91 160 2307845 2 626 3905 3145 1 26 0 73
1 38 1051707 0 0 6 64 16 2313110 2 640 4006 3088 0 26 0 73
1 39 1051816 0 0 3 61 53 2307261 2 676 4246 3171 2 26 0 73
1 41 1051837 0 0 0 0 71 2314504 3 617 3842 3079 1 25 0 74
1 41 1052032 12 0 4 67 64 2310183 2 648 4061 3195 1 26 0 73
kthr memory page faults cpu
----- ----------- ------------------------ ------------ -----------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
1 42 1052032 10 0 0 0 0 2317324 2 624 3917 3077 1 25 0 74
1 48 1052032 3 0 0 0 0 2317809 3 634 3989 3086 1 26 0 74
1 53 1052032 2 0 0 0 0 2318350 2 625 3965 3084 0 26 0 74
1 58 1052032 1 0 0 0 0 2317301 2 630 3920 3089 1 25 0 73
1 59 1051971 150 0 5 48 2 2315600 3 638 4151 3084 1 25 0 74
1 59 1051971 157 0 0 0 3 2317086 2 639 3944 3087 0 25 0 74
1 57 1052008 218 0 4 51 6 2314955 2 642 3969 3104 1 26 0 73
2 55 1052008 217 0 0 0 0 2318663 3 639 3916 3069 1 26 0 73
1 56 1052008 217 0 0 0 0 2317988 2 624 3882 3050 1 25 0 74
1 55 1052259 0 0 7 17 15 2311838 3 670 5057 3125 1 26 0 73
1 56 1052259 0 0 0 0 0 2318664 2 636 3946 3097 1 25 0 74
1 60 1052259 0 0 0 0 0 2317851 2 635 3983 3110 1 25 0 74
1 60 1052305 0 0 5 27 8 2315742 3 656 6059 3194 2 26 0 73
1 61 1052305 0 0 0 0 0 2318552 2 642 3960 3045 1 25 0 73
1 62 1052441 0 0 7 50 214 2307807 2 653 6913 3139 1 26 0 73
1 64 1052471 0 0 6 16 0 2315289 3 651 4212 3155 2 26 0 73
1 65 1052486 0 0 0 0 89 2313722 2 633 3995 3208 1 25 0 74
1 67 1052627 0 0 2 79 17 2310373 2 649 3985 3254 1 26 0 73
1 67 1052653 0 0 0 0 75 2315828 3 649 3962 3156 1 26 0 73
1 68 1052658 0 0 0 0 16 2316820 2 635 4012 3088 1 26 0 74

# ps -aef | grep pmon
oracle 876634 1 0 08:43:27 - 0:00 ora_pmon_VIS
maxcora 1630376 1 0 06:29:33 - 0:05 ora_pmon_MAXC
root 2097198 4808812 0 15:45:19 pts/5 0:00 grep pmon
maximo 3604706 1 0 19:03:02 - 0:14 ora_pmon_maxdemo
crptora 4751424 1 0 06:42:45 - 0:05 ora_pmon_CRPT

$ ps -ef | grep defunct
maximo 3579978 2474170 0 16:22:13 pts/5 0:00 grep defunct
oracle 4022376 1908862 6 0:00 <defunct>
oracle 4489296 1908862 6 0:00 <defunct>

# ps aux | more
USER PID %CPU %MEM SZ RSS TTY STAT STIME TIME COMMAND
root 16388 18.3 0.0 40 17048 - A Oct 17 33777:47 wait
root 40970 18.1 0.0 40 17048 - A Oct 17 33521:54 wait
root 24582 17.9 0.0 40 17044 - A Oct 17 33017:17 wait
root 32776 17.8 0.0 40 17048 - A Oct 17 32999:25 wait
crptora 2580600 6.5 1.0 58252 31924 - A 15:26:31 5:05 oracleCRPT (LO
C
oracle 4939850 4.6 0.0 32176 19676 - A 13:22:17 26:21 oracleVIS (LOC
A
oracle 4563040 2.7 1.0 31504 21388 - A 13:21:25 15:45 f60runm webfil
e
root 57358 0.5 0.0 40 17040 - A Oct 17 1010:04 lrud
oracle 4726976 0.4 0.0 26316 16648 - A 13:22:54 2:04 f60runm webfil
e
oracle 3162182 0.3 1.0 31376 21616 - A 13:52:38 1:34 f60runm webfil
e
crptora 5308466 0.3 1.0 67592 41740 - A 06:43:04 6:53 ora_j000_CRPT
crptora 4956400 0.3 1.0 58396 32128 - A 14:16:03 1:06 oracleCRPT (LO
C
crptora 606330 0.3 1.0 54008 28064 - A 14:05:19 1:07 oracleCRPT (LO
C
root 1163446 0.3 0.0 204 17188 - A Oct 22 411:11 nfsd
oracle 3522724 0.3 0.0 26252 16552 - A 14:12:44 0:58 f60runm webfil
e
oracle 2621652 0.2 0.0 30484 18024 - A 13:23:17 1:19 oracleVIS (LOC
A
oracle 3252416 0.2 0.0 30900 18392 - A 14:13:18 0:48 oracleVIS (LOC
A
oracle 1663114 0.2 0.0 30976 18612 - A 13:53:12 0:52 oracleVIS (LOC


# iostat 2 10

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.1 151.8 21.5 6.8 64.8 6.9

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 2.3 22.5 2.6 10433931 51872839
hdisk0 2.5 26.1 3.2 20278362 51891687
cd0 0.0 0.0 0.0 0 0
dac0 0.0 520.4 254.0 115147211 1326362846
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 20.8 520.4 254.0 115147211 1326362846

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.0 593.4 0.9 25.3 0.0 73.8

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 100.3 198.4 9.2 48 512
hdisk0 100.3 9.9 2.5 28 0
cd0 0.0 0.0 0.0 0 0
dac0 0.0 55.3 13.5 140 16
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 100.3 55.3 13.5 140 16

tty: tin tout avg-cpu: % user % sys % idle % iowait
0.0 435.5 1.1 25.8 0.0 73.1

Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 100.0 116.0 12.5 88 144
hdisk0 100.0 256.0 11.0 24 488
cd0 0.0 0.0 0.0 0 0
dac0 0.0 182.0 43.5 340 24
dac1 0.0 0.0 0.0 0 0
hdisk2 0.0 0.0 0.0 0 0
hdisk3 0.0 0.0 0.0 0 0
hdisk4 0.0 0.0 0.0 0 0
hdisk5 99.5 182.0 43.5 340 24

# sar -u 2 20

AIX hiltrn 1 5 000893DA4C00 11/18/03

15:57:49 %usr %sys %wio %idle
15:57:51 2 28 70 0
15:57:53 2 27 71 0
15:57:55 2 27 71 0
15:57:57 3 25 72 0
15:57:59 1 27 71 0
15:58:02 2 27 71 0
15:58:04 2 28 70 0
15:58:06 2 27 72 0
15:58:08 1 27 72 0
15:58:10 1 27 72 0
15:58:12 2 27 71 0
15:58:14 2 28 70 0
15:58:16 1 28 71 0
15:58:18 1 27 71 0
15:58:20 3 28 69 0
15:58:22 2 27 71 0

#lsps -s
Total Paging Space Percent Used
12288MB 3%

According to the output system seems to be a CPU Bound system.
Bi how many oracle instance we can run on server with 4 GB of RAM and 4 way CPU(600Mhz).the server comes to its normal state if we shutdown the two oracle instances on server.The Server works perfectly normal with two oracle instances.The slowdown behaviour is for a period of 40 minutes in evening hours now.
I have already done some tuning using vmtune.bi i need ur comments on my outputs.

Thanx and Regards

arvibm

 

[gileb]

just do a filemon -o /tmp/file.out
wait for a few seconds and do a trcstop
and show the output in the forum.

 

[bi]

In theory, a 6H1 with 4 GB can handle four databases ... but it depends on the size of each of the databases and the size of the SGA the DBA has set up for each of the databases. How large is this database? How many people accessing it?

I am still concerned that it appears the entire database is on hdisk5. Notice how high your iowait is in you last post and the fact that the only other disks having activity are your rootvg disks. Can you tell us on which disks the database is on? And what else is on the disks?

 

[sbix]

the lrud process is part of the paging replacement mechanism, probably you have an heavy pagination problem.
If so it is not so strange that you are experiencing connection problem, cause each new connection means new processes and therefore new memory request

 

[arvibm]

Hi,

The entire database is on hdisk5 (array of 7 disks-raid-5).The database is accessed by around 200 users.Rootvg is on hdisk0 and hdisk1(mirrored,but paging space is not mirrored).No data files or anything related with database is on rootvg.hdisk 0 and hdisk1 only contains rootvg.Other disks hdisk2,hdisk3,hdisk4 belongs to different server.we have implemented HACMP between database server and training server.
hdisk2 ,hdisk3 and hdisk4 belongs to database server.I am facing problem with training server only.The database logical volume size on training server are:-
#df -k output
/dev/crptdata_lv 67108864 20680708 70% 82818 1% /crptdata
/dev/maximo_lv 26214400 5161284 81% 184151 3% /maximo
/dev/devl 68157440 8304344 88% 284345 14% /devl

I am still searching if the problem is with server memory size/tuning or problem related with database tuning.according to my view if the problem is with operating system / server then the problem should remain throughout the day.The server goes slow down for a period of 40-50 minutes.The server works perfectly normal with two oracle instances but the moment i start two more oracle instances the free ram pages value goes to zero.
gileb :-i think the filemon output is useful only at the time of problem.i will try to collect the output of filemon tommorow evening.
sbix :-what is the solution for heavy pagination problem?

regards

arvibm

 

[bi]

I think the reason why everything slows so much when you start the other two databases is because the total SGA for all the databases that are running exceeds (or almost exceeds) the memory of your system. Ask your DBA to give you the Total System Global Area size for all four of the instances (SGA is given in bytes), add them up and see what the total is. If it's close to your memory, that is part of the slowdown.

Has the DBA put the instances in archivelog mode? I was surprised when a DBA I was working with put a large database into archivelog mode and he got faster performance. I think it is because when the logs had to switch to a new file, the database didn't have to wait for the file they were being switched to to be deleted.

Here are the vmtune parameters I used on an H80 (the predecessor of the 6H1) with 8 GB of memory and 6 CPUs (@ 500 something MHz, I think), and SSA disks:
/usr/samples/kernel/vmtune -c 2 -R 1024 -f 320 -F 1344 -W 512

The -R parm is quite a bit larger than what you have. Remember, though, that changing vmtune parameters takes time and working with the DBA to measure if the changes are having the effect you want.

I'm still concerned that the entire database is on one RAID5 array. Although some will argue that with the current disk technology it doesn't matter and it seems that more DBAs aren't separating their files from each other (a result of folks who don't really understand LVM), just about everything I have seen from Oracle and IBM says to:

1. avoid RAID5 for write intensive files. In an Oracle database, you have logs and rollbacks that are very write intensive. When each one of those 200 users makes any change in the database, you've got logs being written and rollbacks being written until the user's change is committed. With RAID5, a write to the disk isn't completed until there are at least two reads and two writes of the disk. This will cause a delay (slight, but still a delay). Do you have write cache enabled on the array? That will help with the delay and IBM says you may get as good performance is for RAID1.

2. separate write intensive files like rollbacks and logs onto separate disks and through separate controllers if possible.

3. Don't put the indexes for your datafiles on the same disk as the datafiles.

Even with 7 physical disks in this array, you have no control over where the data is. So, your seek times may be quite large.

This array is on one controller, right? So you've got all this data going down one pipeline. That contributes to slowdown.

I don't know if you have read this RedBook on the FasTt500, but you might find it useful:

http://www.redbooks.ibm.com/redbooks/pdfs/sg246246.pdf

Also, at the RedBooks site, take a look at Database Performance Tuning on AIX (Pub # SG24-5511-01).

 

[sbix]

You're experiencing problems dued to lack of memory.
There are two ways to face this problem without adding more physical memory.
DBA can use raw devices instead of files on filesystem for the Oracle datafile, or can look the DB_CACHE performance in order to see if a reduction is possible.
A significant reduction of memory use can be achived by using the MTS (Oracle Shared Server)

 

[arvibm]

Hi,

Thanx for all the replies.i have forwarded ur database related tips to my DBA and he is taking help from oracle support to tune oracle database parameters.
We have noticed a database process consuming around 25-30 % of cpu and when we checked for that particular process using &quot;oradebug setospid <process number> &quot; on sql prompt it shows the process is TNS process.
we have checked for other process consuming CPU more then 10-11% and all the process are TNS process.
According to my view the TNS process is for network requests.
Is there any way to find out why the TNS Process consuming high percentages of CPU.?

Regards

arvibm

 

[arvibm]

Hi,

Thanx for all the replies..finally decided to go for more memory.

Regards

arvibm