You folks are freaking me out. The OP could easily be handled by testing with a scan "agent" that doesn't vary its I/O profile. I said so straight out of the gate and Mark said so with very rich supporting text. Let's just see what the array head can deliver to dd(1) since it scans with precisely the same sort of I/O as a scattered read except (most critical) it is buffered in the data segment of dd(1) instead of sprayed across buffers in the SGA. ok, crawling back into the cave to work on the blog kevinclosson.wordpress.com ________________________________ From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Vlad Sadilovskiy Sent: Sunday, December 10, 2006 12:03 PM To: Mladen Gogala Cc: knecht.stefan@xxxxxxxxx; Hemant K Chitale; oracle-l Subject: Re: I/O and db_file_multiblock_read_count And what did I say that was different from your first sentence? Second sentence I think is wrong. I'll deliberately use explicit tag for parameter and system statistics MBRC value to reduce possible misinterpretation.