UNDO retention causing excessive I/O ?

Question

New DBA here.
Quick Summary:
We have an oracle 10gR2 DB on AIX 5.3. OS runs from the box, Database runs from a san. Aside from the DB, there is a web based application running on the box, which uses the DB.
System is dedicated to this application.
We have a consulting company that is making suggestions, and I am not sure about some of them.
When we set up the software, the vendor told us to set up the database with undo retention to 10800. So we did.
Our consulting group is telling me that Undo I/O represents 34% of the total database I/O, and that I should reduce the retention time to 1080. *(they also tell me that the longest running query is 100s; I don't quite understand how this is relevent).
Why would reducing the UNDO retention lenght reduce the UNDO I/O ?

Answer 1

First you need to check the concept of UNDO, the main purpose of undo are:
Rollback an active transaction
Recover a terminated transaction
Provide read consistency
Recovery from logical corruptions
http://download.oracle.com/docs/cd/B19306_01/server.102/b14220/logical.htm#sthref449
also check undo retention,
The undo retention period indicates the amount of time that must pass before old undo information—that is, undo information for committed transactions—can be overwritten.
As you can see, there's no direct relationship between undo retention and database I/O. Reduce undo retention will not reduce database I/O.
Besides, the UNDO_RETENTION initialization parameter is ignored unless retention guarantee is enabled.
*(they also tell me that the longest running query is 100s; I don't quite
understand how this is relevent).The relevant here is your undo retention should be longer than longest running query (and undo space is big enough to hold the undo data of longest query) , so that you don't have out-of-space conditions in the undo tablespace and ORA-1555 error for consistent reading.
However again, it's relevant to undo setting not database IO.