Recommended init.ora values for PROD setup

Question

Hi All,
-Assuming that I am using a 64-bit version of Oracle,
-that there is nothing else running on the server,
-that there will be no other databases on the server,
-that I have substantially more data on disk than I have RAM,
-and that I am using 11g. I want to setup a production database 11g. The server specs is OEL 5.6 with 12Gb Ram and a Quadcore CPU.
There will be 200 users simultaneously accessing the database. I understand that 10g and 11g are "self-tuning".
So in this regard do I need to still change values in the default init.ora?
Can you give me sample PROD template of your init.ora?
Or maybe since my RAM is 12Gb I will just set the "target SGA = 8Gb" and the database will now be good?
Thanks,
Yx
Edited by: yxes2013 on 8.2.2013 16:21
Adding Justin's assumptions

Answer 1

I Thank you all :)
You need to understand what the memory is used for. Please read the documentation on SGA and PGA components and parameters and how they impact performance.Yeah I read it already....but it does not clearly explain like what all you did here ;)
Hmm interesting....I got lots of tips learned. So If I have 500Gb data (im sure it does not include index right?) Broken into the following:
Table1 300Gb
Table2 100Gb
Table3 50Gb
Table4 30Gb
Table5 10Gb
Table6 5Gb
Others 5Gb ( comprised of hundreds of tables)
Do I need to have my memory target SGA equal to 500Gb to attain a maximum performance? supposing memory is not a budget constraint?
Supposing Table6 is the most frequentlly read/accessed and I want to load/cache it all in the memory so I do not need to index this table, how can I do it?
What is the rule of thumb of PGA size against the size of the table? Maybe 20% sorting of the most accessed table?
@Justin
**10 sessions doing data warehouse queries will probably need more PGA than 1000 sessions doing OLTP queries.
**I'd probably start with a MEMORY_TARGET of 10 or 11G. How about the approx. PGA size dear? :)
@Hemant
It is more the amount of data and some factoring for the number (volume) of changes to the data.
The number of users matters in terms of sizing memory for server processes, and some allocation of the SHARED_POOL_SIZE, not for the DB_CACHE_SIZE component.
If more users means more distinct SQL statements, the SHARED_POOL_SIZE usage would be greater.
However, multiple users accessing the same version of the data will be "sharing" one buffer in memory for each block of data being accessed. It doesn't matter if 1 user is reading a block or 8 or 64 or 128 users are reading the same block -- in either case it is one buffer in memory.
It is only if the block is frequently being changed and concurrently queried, might there be current and a few consistent versions of the buffer being maintained in the SGA.I think in 11g I do not need to set those because it is "self-tuning" right? All I need is the target SGA size. :)
Given that I will perform a "benchmark" test/simulation on the DEV server. Is there a tool that will gather the right sizing of init.ora parameters? Can I extract it from AWR report?
Thanks to all again,
Yxz