Hi,

We run Oracle 12c RAC with IBM AIX 7.1 TL5. We got some performance on application becasue some our DBA ran oracle advisor and it blocked another oracle process. Oracle supporter asked our DBA ran Oracle OSwaatcher and then Oracle supporter said with our DBA this problem came from the system has high context switching cause from high memory scanning in Oracle. I donot agree with them because 

1. high memory scanning report should be in memory occupied by Oracle. not OS.
2. I told them high context switching should relate high CPU usage instead high memory
3. from OSwatcher report, I saw OSwatcher report and I saw pi/po, fr/sr always are 0 but has fo in 3 digits (about 100-110), and saw a lot of Oracle threads in top, not LRUD process ran. 

I do not think high context switching cames from memory thrashing. 

Am I correct? And any recommendation?

Thanks

Kritsada.

I am not an Oracle guy, but because you ask for recommendation - may be it is time to upgrade to AIX 7.2? or 7.3? And afair Oracle has a newer version too. Something like 19c.

Everything else very relative to your system. What is "high context switching"? Do you have any numbers? How is your server dimensioned? Is it SPLPAR or dedicated? How many CPUs and RAM does it have? Is SMT on or off? There are so many settings which can change Oracle behaviour on AIX that I don't really know where to start ;-)

What I can say, if pi/po in vmstat (I don't know OSwatcher) constantly show 0, it is a good sign. fr/sr I'd say normal operations of LRUD and if you see it on top or not, doesn't really matter.

Hello Kritsada,

It is very difficult to troubleshoot performance issues like yours just on the basis of your posted details. Do you have any PMR opened with IBM with more information?

I work in IBM Expert Labs and we do this kind of troubleshooting and performance tuning assessments regularly for Oracle workloads on AIX. If your company has purchased any IBM Power systems in the last year, there may be Expert Assist Reward points allocated to you. You can utilize these rewards points to set up an engagement for us to review and troubleshoot your system.

Send me an email and we can figure out what can be done.

sujit.dutta@ibm.com

------------------------------
SUJIT DUTTA
------------------------------

I would suggest following the IBM support procedure. I believe they will ask if the context switching is on the LPAR or the overall box. Entire system lag due to high context switching on an entire box can occur on a busy system with a high ratio of virtual processor to entitled capacity without weights. This is rare but it depends on how things are configured and running. Perhaps bring this up with IBM support if you feel they have not explored this possibility.

P.S. Upgrading to AIX 7.2 or 7.3 is recommended but won't necessarily resolve this issue.

------------------------------
Jeff Boleman
------------------------------