There was a thread about FFS some time back. It turns out that ext2
is better suited to the x86 architecture than the BSD FFS code. The
ideas of FFS are implemented by ext2, but are done a little better if
you believe the ext2 developers. It is hard to make any concrete
claims about the filesystems since the real bottleneck has little to
do with the filesystem structure. Once we got clustered reading and
writing and once we got distributed metadata, the real problems are
parallelism and spindle speed. A project at UCB called XFS aimed to
solve this by distributing metadata and between networked hosts. It
is a clever idea that has not yet come to fruition.
-- Oscar