diagnosing physical IO issues beyond the database

I have run 10046 traces on full tablescans of the same table in different
databases. The overall physical IO response time and the db file scattered
read waits are up to 12 times longer than in others. When I go to the raw
trace and look at the ELA per call, I see the same thing.

We are on a SAN. Our SA has looked at the "disks", probably with IOSTAT and
does not see anything. To Oracle this is a physical IO issue, however,
underneath it could be alot of other things. I am not proficient at
diagnosing at the OS level or lower. Does anyone have suggestions as to
where we should start looking?

Other related posts: