HP-UX 11.00 and Performance 1

[ScottAB]

We are running an application written in Forte that runs slower while compiled on our HP-UX 11.00 box than it does in interpretive mode under Windows! This just doesn't seem possible so we are looking into fine-tuning how Forte / HP-UX should be configured. One item we found is to use the ndd command to bump up the tcp_conn_request_max parameter for /dev/tcp (Forte's recommendation). Another idea is to increase the number of threads available for a process. Is there a way to this? Is there something else that we should be looking at? Any input will be greatly appreciated.

 

[crowe]

there is a kernel parm you can change called max_thread_proc. this is configurable in sam.

crowe

 

[ScottAB]

Will changing this parameter require a reboot?

 

[reply]

we need to know how the details of preventive maintaince of all hp-ux system. becoz ours is a big project where we had installed many system of this.
o.k
we wish that we get positive reply.
bye..

 

[skyl1nev6]

This change will require a reboot. Matt
Sr. Systems Administrator

 

[jollyroger]

Are you running any other software besides forte on your machine such as Oracle?
Regards
Ben

 

[ScottAB]

Yes, we are running Oracle 7.3.4.5, 8.1.5, and 8.1.7 on this box.

 

[MikeLacey]

That's potentially quite a busy machine

- what does 'uptime' report?
- does 'vmstat 1' have values in the pi or po fields?
- does 'iostat 1' show an overloaded disk subsystem?

Mike
"Experience is the comb that Nature gives us after we are bald."

Is that a haiku?
I never could get the hang
of writing those things.

 

[jollyroger]

Hello there
We have the same problem and it relates to the Context Switching on the machine.
Forte and Oracle together require a lot of attension from the CPU, and thus fight for its use.
If you use the command vmstat 5 5 on a 132 width screen, you will see the column cs for context switch.
If you stop and start either of the two applications forte and or oracle, you will notice this drop drastically.
My suggestion is to either get another CPU installed or you can further look into it by using Glance to monitor and identify which process is causing the most amount of context switching (i.e. a process which is taking time to finish etc). Have a look at this article:

http://searchhp.techtarget.com/ateQuestionNResponse/0,289625,sid6_cid424645_tax288519,00.html

With regards
Ben