In message <199601182321.XAA05041@doublesix.domino.org>, "Neil J. McRae"
writes :
>Yeah but it seems FFS itself is slow, is anyone doing anything with LFS?

I don't know what the current status is, but about the FFS;
has anyone looked into turning _off_ the kernel reordering of writes?
I read in the ATA/EIDE faq (a very good reference for ATA drives) that
the "reported" geometries are not indicative of internal geometries -- this
may also be true of SCSI drives (I'm not sure).  In this case, the kernel
would be slowing down by bothering (and could even be detrimental by
reorganizing in a bad order).
Meanwhile, I hear that some new drive/controller combinations do some of
their own reordering (elevator?  scan?) and thus kernel facilities would
be redundant and slow.
Finally, some drives have variable density cylinders (zones, where the #
of sectors per cyl varies from zone to zone).

Are there defines to turn off this behavior?
Know how to get the true internal geometry? 
Is the SCSI reported geometry during autoconfig the true internal geometry?