CPU / Memory Upgrades...

[Shaun Aldrich]

Hello Everyone,

This past weekend we had just completed a major upgrade where we added more CPU / Memory to our K570 & K210 systems. These systems are running HP-UX 10.20.

We are running Oracle 8.0.6 on these systems as well. If you look at the CPU Avg load as well as the idle time, you will see some great readings. Even Chapter 3 which is the hardest hit, still maintains lower than a 2 for the average load and actually records some CPU idle percentage.

This leads me to the following few questions:-

1) What is the highest the CPU Avg load should read before something needs to be done?

2) Now that the Hardware Upgrades are complete, does our DBA need to adjust any Oracle parameters to take advantage of the additional memory?

3) How could I find out what the recommended kernel parameters should be adjusted to take advantage of this CPU / Memory increase.

Thanks for any help or advice anyone has.

S Aldrich

[Rick Garland]

1) The highest load you want to see is dependent on your environment. A load of 2 on a system with poor response time is a system that needs to be attended to. A system with a load of 10 but provides the necessary response times as required is OK. load values are dependent on your environment.

2) Check with the DBAs. There will likely be a desire to increase the oracle SGA as a result of increasing the resources.

3) Most of the kernel changes/mods will be from the DBA side.

[James R. Ferguson]

Shaun:

Your load average isn't bad. The real measure is the response your users see. A number of very good suggestions appear in an earlier post. See:

http://my1.itrc.hp.com/cm/QuestionAnswer/1,1150,0x727e0559ff7cd4118fef0090279cd0f9,00.html

Good luck.

...JRF...

[Sanjay Dudeja]

Shaun,

To tune the kernel you can - you can compare your current kernel parameters against the tuned parameter set provided by the system. To view this you need to go through SAM-> Kernel Parameters -> Configurable parameters -> Action ->Apply Tuned parameter set.
All Semaphore, Shared memory related and Message queue related paramters need special attention in a RDBMS environment.
Oracle Shared pool size needs to be increased in init.ora for taking the memory advantage.

[Mark Mitchell]

I can't tell you what is a good load because that is realitve to your enviroment and how the users feel "often the best source". What I can recommend is a book put out by HP labeled Performance and Tunning which will show the methods and planning required to help always be ahead of the curve.

[CHRIS_ANORUO]

Hi Shaun,
Take a look at this problem from the application angle. Find out from program "top" what processes are acually doing all the work. Use Glance and check out the wait states that these process are consuming.

Is your database using the local UNIX domain protocol or the high overhead TCP/IP to do its local database access? (This is a common problem with local database access.)

Take a look at your buffer cache hit rates if you are doing filesystem I/O. Use sar -b for this.

How much time is being spent in user mode vs kernel mode? Check this with sar -u.

Also, check the %wio parameter from sar -u. This may lead you to an I/O bottleneck.
Use sar -d for disksubsystem analysis
Use sar -M for multiple CPU's
Read more about sar if you don't have glance+ installed on your server.