[ Friday, March 3, 2000 ] Ricky Beam wrote:
> As I understand it, the "stride" will only make a real difference for
> fsck by ordering data so it's (more) evenly spread over the array. This
> sounds correct and even "looks" correct when observing the array -- but
> I've never bothered to look at the file system handling of striding.
I'd always imagined it allowed the ext2 layer to aggregate data blocks
(the number to aggregate being the stride param) before passing the
blocks to the md layer, making things more efficient since the md layer
wouldn't have to do the same aggregation and could simply pass down
a single block. Nothing based on looking through code, just an impression.
It'd be good to know, actually :)
Seems like you'd ideally like ext2 to pass down the data in full-stripe
sizes, but that could be asking a bit much.
James
