RE: I/O and db_file_multiblock_read_count

 >>>I'm still thinking that the data was coming from array cache 
>>>though, but I may have tangled up service requests for i/o 
>>>with the blocking requested by dd, and that may not be 
>>>identical in all the layers down to the rust.

yeah, I'm with you, Mark...it is such an astonishingly small
datafile it is cachable in a wide variety of places...to include
what representation it has in the 8MB track buffers there is on
all the involved drives...and I don't think we've lokoed at whether
this is even direct I/O.

What I don't want to see happen is some carry-away notion that
larger I/O requests are slower... yes, sometimes man bites dog :-)

--
http://www.freelists.org/webpage/oracle-l


Other related posts: