How to increasethe throughput of the I/O subsystem in Oracle 10g?

Question

How to increase the throughput of the I/O subsystem in oracle 10g.
OS : Windows 2003 Server

Answer 1

I think your ADDM report may well be wrong.
In another thread you've reported the ADDM for this 23 hours as showing::
This is my ADDM report:
Analysis Period: from 06-MAY-2008 09:30 to 07-MAY-2008 08:30
Database Version: 10.2.0.1.0
Snapshot Range: from 2656 to 2679
Database Time: 100249 seconds
Average Database Load: 1.2 active sessions
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
FINDING 1: 100% impact (100249 seconds)
Significant virtual memory paging was detected on the host operating system.
RECOMMENDATION 1: Host Configuration, 100% benefit (100249 seconds)
ACTION: Host operating system was experiencing significant paging but no
particular root cause could be detected
At the same time, your extract from the matching AWR report shows:
Event Waits Time(s) Avg Wait(ms) % Total Call Time Wait Class
db file scattered read 20,665,706 48,083 2 48.0 User I/O
CPU time 30,114 30.0
db file sequential read 3,661,225 14,740 4 14.7 User I/O
read by other session 454,471 2,933 6 2.9 User I/O
db file parallel write 485,625 1,426 3 1.4 System I/O
So the AWR in these 5 events shows 97,000 seconds of time, but 100% of the total DB time is supposed to be related to Host swap time. The two sets of numbers are not consistent.
Moreover, the wait times are showing "db file scattered reads" with an average time of slightly less than 2.4 milliseconds. It would be a little surprising if you have a swapping problem AND I/O times that were operating at cache speeds. (Your db file sequential reads also have a very good average wait time).
23 hours is not a helpful interval though - the average may be hiding shorter periods of nasty activity when reponse time is bad. So (as Charles Hooper indicated) - where, or when, do you perceive a performance issue; and what do you get from ADDM or AWR for a shorter time-range around that period.
My first thought, looking at the numbers you've supplied, is that if you think some critical SQL is running slowly is that you're probably working too hard because you are missing (or not using) some critical indexes. In your place, I'd run off a few one-hour AWR reports in that 23 hours, and look at 'SQL ordered by elapsed time' or 'SQL ordered by reads' for the top few statements, then pick up their execution plans by running the awrsqrpt.sq report.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
http://www.jlcomp.demon.co.uk
P.S. Please check the FAQ (see top right of page) for details on how to use the PRE tag to supply code and reports in a fixed font that makes them more readable.