This is just a guess - I have significant experience of Scsi drives but none of Unix
internals. To
do a good sync, you have to force the data from the CPU to the disk, and from the disk
ram onto the
disk oxide. IDE disks are not very clever, and I do not think that they cache
unwritten data. If,
therefore, the data has left the CPU, it will have hit the oxide. Scsi disks, however,
play
considerable tricks, which may include delayed writeback. If I were writing a Scsi
disk drive, I
would be strongly tempted to put a Scsi Rezero command into the sync operation. This
has the effect
of flushing all cached data to disk - amongst other things.
It is the "amongst other things" which is the catch. Some disk manufacturers just do a
simple reset
of the disk's seek logic, which would only take a few milliseconds. Others treat a
Rezero command as
an instruction to do a full thermal recalibrate, which may take 250 milliseconds. This
means that
drivers tested on one brand of disk will show essentially no performance hit from
doing a sync with
Rezero, whilst a different brand would show a collossal hit. Yours appears to fall
between the two
extremes.
Alec Cawley
---------------------------------------------------------------------
Before posting, please check:
http://www.mysql.com/manual.php (the manual)
http://lists.mysql.com/ (the list archive)
To request this thread, e-mail <[EMAIL PROTECTED]>
To unsubscribe, e-mail <[EMAIL PROTECTED]>
Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
