Currently the alignment is based ONLY on the physical
start of disk/file. There is no way to "slide" the aligment
Say you are testing RAID5 (64k stripe), and your file is
not aligned on the stripe size, a 64 K read will involve 2
drives. There is no way to align the read requests
properly to hit only 1 drive.
The logical solution would be to have the aligment be
slided with the starting sector number. That way we
could align the reads properly.
Another more complicated enhacements is when you
want to test raid5 hitting the same drive each time.
Because of parity, in adition to aligning on the stripe size
* (number of drives-1) you have to skip an alignment
every N alignements, where N is the number of drives.