Oracle Database Discussions

Hi.
Firstly environment:
Windows 2003 Server Enterprise Edition, Service Pack 2. 3.5GB RAM.
Two Oracle Databases running on this server, one not heavily used.

On the database i am concerned with it is 10.2.0.3.0.

I have just started to look at this database and am seeing various things in the alert log and have had user error messages as well.
Initially when i started looking at this server it was running with quite a small SGA (600MB) and the memory_target_advice was saying to take it to 1000MB.
(PGA is around 200MB)

In January, before i looked at it, it started to throw:

ORA-04031 unable to allocate 16 bytes of shared memory ("shared pool","........", "sql area", "tmp")

These would occur regularly and mostly during nightly processes, it seems, and thus not bubbling up to users.

Having switched the SGA_MAX_SIZE to 1GB a new set of errors starts to appear in the alert log:

ORA-27301: ..... Not enough storage is available to process this command
ORA-27302: ..... ssthrddcr

At the same time the ORA-04031 errors start to bubble through to the users.
Before i could return to the scene of the crime someone else set the SGA back to 600MB and the error messages stopped. However the ORA-04031 message start to reappear in the alert log.

My questions are many and, mostly, from a lack of knowledge ;) But some guidance on the above would help me in not reading way too many forums.
In particular:
1. Why would increasing the SGA lead to a different memory error? I would assume this would have something to do with the amount of memory a 32bit OS will allow a single process to address but the PGA+SGA is only 1200MB.
2. How does a ORA-273xx error end up with a ORA-04031 being shown to the users? I would look at the listener log but its over 2GB!!!


Thanks in advance for any advice.