Your tests prove that the I/O subsystem drops from 115MB/s to 40MBs if doing a sequential scan of a really small file. It generally is not the case that larger application requests result in longer scan times, but it raises some questions: 1. Is users01.dbf the same file invovled in the test reported in the OP? 2. What kind of SAN is under users01.dbf? 3. What kind of LUN is under users01.dbf (e.g., 2 disks RAID 1+0 striped with 32KB ?) >>>-----Original Message----- >>>From: oracle-l-bounce@xxxxxxxxxxxxx >>>[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Kevin Lidh >>>Sent: Monday, December 11, 2006 6:39 AM >>>To: Kevin Closson >>>Cc: oracle-l >>>Subject: RE: I/O and db_file_multiblock_read_count >>> >>>I apologize for the tardiness of my response. This is the >>>result of timing the dd. It seems to indicate the same >>>thing as the scattered read test. Or am I reading that wrong? >>> >>>(8k * 16) >>>Unix> time dd if=/ora/data/bordcdw/users/users01.dbf of=/dev/null >>>bs=131072 >>>1290+1 records in >>>1290+1 records out >>> >>>real 1.4 >>>user 0.0 >>>sys 0.5 >>> >>>(8k * 128) >>>Unix> time dd if=/ora/data/bordcdw/users/users01.dbf of=/dev/null >>>bs=1048576 >>>161+1 records in >>>161+1 records out >>> >> -- //www.freelists.org/webpage/oracle-l