Blguy,

Interesting Topic...........

The following suggestions may help you:

1. Use RAW devices for all oracle files.
For > 1 TB, I would always use RAW DEVICES with (RAID1+0).

2. Use the DB server's internal disks for the redo logs and reduce the redo buffer size to 1 or 1.2 MB.
Or use raw devices atleast for REDO LOGS. Defintely this will reduce "log fule sync" waits.

3. Ratio pf Phy.Read to Log. Read
From you statspack report:
Logical reads: 16,930.78 184.73
Physical reads: 2,810.45 30.67

The ratio of phy to log is 16.59 percent.
This seems to be high (how high I leave it to you).
In general I expect this value should be around 3 to 5 percent.
This indicates the more physical IO calls are issued by the system because of the data blocks not available in the SGA.
Try to increase db_block_buffers parameter.

How many "user commits" were executed during this 10.27 min?
This statistic is in the statspack report.

If possible, you can attach the entire statspack report in this thread.
That may be helpful for all.

By the way, did you tune all TOP-N SQL statements?

If you say the SQL code cannot be modified, I will not accept it.

I have modifed Siebel/Peoplesoft/Oracle CRM codes in the past 5 years.
Those big companies all accepted my changed codes.

Tamil