We have a customer-managed Oracle instance running on EC2 server and I want to run the DBMS_RESOURCE_MANAGER.CALIBRATE_IO package procedure to obtain better estimates of disk i/o. We have a query that is invoking 12 parallel threads. The parameter num_physical_disks has a default of 1; which I assume would be a horrible default for this. I'd guess I would use maybe 10 for a wild guess. I am new to the AWS world, so I don't know how our Linux mount points are set up. Here is my /etc/fstab for /u01 and /u02:
/dev/mapper/atlas_vg-u01_lv /u01 xfs defaults,nofail 0 0
/dev/mapper/atlas_vg-u02_lv /u02 xfs defaults 0 0
I have run the DBMS_RESOURCE_MANAGER.CALIBRATE_IO package procedure several times, with various numbers of disks, such as 4, 10, 20, and 30. The results from the output of those runs show very neglagble difference. So pretty sure it would be safe to use 10 or 20 as a value for the num_physical_disks parameter.