This one's a bit more specific than the last. If memory serves me,
XFS supported differing block sizes (which you could do on a hardware format
of a SCSI drive) back on IRIX. But when first ported to Linux it didn't
work. Was that ever fixed? I seem to remember that going to 1-2K block sizes
gave and extra 10%, and it almost seems logical that going to a 4kK block size
would be ideal for xfs (presuming your disk doesn't start getting errors,
then it might get harder to remap sectors and you'd hit hard disk failure
(w/o remappable sectors) sooner.
But at least 1K might be a reasonable tradeoff? Been quite a while since
I tried it and don't even know if the SAS drives allow it (if they do, I wonder
if the newer SATA drives do?)
thanks more... :-)
-linda
|