topic Re: performance issue on L-2000 in Operating System - HP-UX

performance issue on L-2000

Deepak Seth_1 — Wed, 09 Feb 2005 13:41:07 GMT

we have l-2000 (4 cpu and 6 gig mem) . dbc_min is 2 and dbc_max is 8 . total swap is 6 GB . iam seeing a memory utilization of 99-100 all the time . i used MWA and grouped the application to different grpup . The RSS total for all application is 2 GB . I also used this command "UNIX95= ps -eo "pid ruser pcpu vsz=Kbytes" -o comm | sort -rk 4,4" to look for processes using more memory . If i total the Kbuytes of this output , i get some weird memory total . I see pageout happening on the system all the time (10-15 nos. / 5 minutes avg ) .

The memory utilization never comes down even during the time when there is less workload on the system .

- Is there any process eating memory ?
- how do i make decision on buying a memory ?
- Should i increase my swap ?

Please suggest .

Re: performance issue on L-2000

Florian Heigl (new acc) — Wed, 09 Feb 2005 13:47:57 GMT

I think this is space allocated is pseudo swap, but I'm not the memory management guru in any way :)

still, have a look at swapinfo for the 'mem' columns.

also 10-15 pageouts per 5 minutes should be way to little for a problem.

Re: performance issue on L-2000

David Child_1 — Wed, 09 Feb 2005 13:56:23 GMT

Deepak,

What application(s) are you running on this server? You should take a look at shared memory (ipcs -ma) to see if this is where your memory is being used. I have run into similar senarios and after looking at shared memory I found Oracle had allocated a bunch of memory. I then found out from the DBAs that they had bumped up their memory settings to "improve performance". Of course in the end it killed their performance.

David

Re: performance issue on L-2000

RAC_1 — Wed, 09 Feb 2005 13:57:19 GMT

What does glance -m says??

What does vmstat 5 5 says??

What does

UNIX95= ps -ef -o "pid,args,vsz"|sort -nk3 says??

Re: performance issue on L-2000

Deepak Seth_1 — Wed, 09 Feb 2005 13:57:48 GMT

swapinfo -amt
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4096 978 3118 24% 0 - 1 /dev/vg00/lvol2
dev 2048 984 1064 48% 0 - 1 /dev/vg00/lvol11
reserve - 2037 -2037
memory 4939 3721 1218 75%
total 11083 7720 3363 70% - 0 -

I won't say it is big problem now but we have some projects coming up soon and i would like to do study on the memory and cpu usuage of different applications .

Re: performance issue on L-2000

Deepak Seth_1 — Wed, 09 Feb 2005 14:07:04 GMT

we have clarify databases with some of them using e-support (JAVA ......)
i have attached a doc with all you wanted . I have gone through the ipcs and couldn't find anything unattached as well checked with my dba on shared pool for oracle . We have 3 main DB and shared pool is resonably small .

Re: performance issue on L-2000

Deepak Seth_1 — Wed, 16 Feb 2005 14:02:55 GMT

i want to update you on fe mor things . I did little study over the weekend.

- i shutdown my DB/APP one by one and check that how much memory laod it reduced .
- After shutting down everything (all DB/APP) and killing all app processes , the memory utilization reduce to 75% .

- In fact there was nothing running at the system and it was still using 76% of memory ?

How can a system be using 76 % when there was nothing running ? I checked any ghost processes using "lsof" and couldn't find anything . Is it possible that memory leak is happening on system processes ?

i rebooted the server and it came up with around 25% and after starting all db/app , it goes upto 47 % . Now on a working day with about 120 oracle connection , it is running around 80% .