You might seriously consider getting rid of AMM and using medium pages. http://public.dhe.ibm.com/partnerworld/pub/whitepaper/162b6.pdf (although link seems broke to me and I've seen it before, it's referred to in many places). Search MOS for how to do it and which patches to get from ibm.
Edit: Oh, and yes, set a minimum shared pool in the meantime, work through 4031 doc on MOS to perhaps fix whatever problem there is.
Before suggesting you change your memory managment in use I would want more information such as how long have your been running the system this way? How long had the system been running since its last startup?
If you have access to the AWR then what does the AWR report for the problem period show? Any unusual activity? Were any new features or major processing added to the application recently.
When you look at the memory management views how frequently does Oracle move granules from PGA to SGA, from the buffer cache to the shared pool and so on.
The following support note may help with trying to investigate what issues you face in relation to ORA-04031 providing the issue re-occurs.
Oracle document# 1088239.1 Master Note for Diagnosing ORA-4031
HTH -- Mark D Powell --
If I've understood well you got a connection problem during 14 minutes due to an ORA-04031.
So may be the Database increased already the size of the shared pool and there's nothing more to do.
To verify that point, you may query the View
Else, as answered before, even if you use AMM, you are able to set a minimum value for the Shared Pool with the parameter SHARED_POOL_SIZE.
Hope this help.
Thank you for your recommendations. I wanted to mark your answer as "Helpful" but was limited to 2 responses for that category. I will definately read up on using medium memory pages for future reference. I can then discuss options with our AIX sys admin.
I appreciate that you took the time to share your knowledge with me.