LID and Portal performance problems on LSF9 on AIX

What values do you have for INSERTBUFSIZE and ARRAYBUFSIZE?

They're not set at all. We went from a single to server to two (one with Oracle only) and the values are not in the ORACLE file. What's it defaulting to? What should we set them to?

I've now found that in LSF9 the default for both of these is 10. So if i don't have them set, they're at 10.

Still, I need to know if the oradb10 processes are behaving correctly. Any more ideas would be appreciated.

WEe might just need more RAM. If anyone out there is running LSF9 with pflow and Oracle on a separate server, I'd love to see the output of this command:

ps avg|sort -n -r -k6

I'd also like to know your processor, memory and swap space configuration. This might help me determine if we have a problem or simply need more RAM. The sizing documents from Lawson lead me to believe we are fine. That's why this is so confusing.

I'd bet you do need RAM. I don't have much confidence in any sizing I have seen from Lawson. But WebSphere uses quite a bit of RAM,

$ lsps -a
Page Space Physical Volume Volume Group Size %Used Active Auto Type
paging01 hdiskpower73 pagingvg_new 8192MB 1 yes yes lv
hd6 hdiskpower0 rootvg 8192MB 1 yes yes lv

We used to be a bit tight on RAM, but since moving Oracle to a separate server and therefore freeing all of the RAM Oracle previously used, we currently don't have any paging. No complaints here about online response. Batch is a bit slower than before LSF, but not seriously so.

Ben,
How much RAM do you have on your app server? Is it running pflow, apps and websphere (everything but Oracle)? We're paging up to 80% in heavy times. We have 8GB RAM and 16GB paging to do everything but Oracle. I want the ps avg output because I'm not convinced that the oradb10 processes are behaving correctly. Plus, there are often over 100 oradb10 processes running.

My guess is that if you had 16 or even 12 you would see a big change in performance.

Thank you. You have 32? How did you land on that? A Lawson recommendation?

I appreciate the input. I'm going to compare the memory stats from your oradb10 to ours.

Ben,
Can I get in touch with you on your database config?  We have our ladb processes down to 18000 each but yours show 3000.  We're not sure how to furhter tune the oradb10 processes.

This is finally fixed. Our DBA went with his hunch and installed the full Oracle database with client. He then moved the client to ORACLE_HOME and deleted the database. This installed a "more complete" client. I'm still unsure of the specifics. I'm discussing details with him this week.

Now processes are miniscule. For future reference, it's all about the client. We had three different client attemps and apparently we needed the most robust one.