Pluggable Disk Schedulers in GEOM
nospam at nospam.com
Fri Jan 5 21:52:24 UTC 2007
>From Luigi Rizzo <rizzo at icir.org>, Fri, Jan 05, 2007 at 10:29:05AM -0800:
> Essentially, a process doing e.g. a 'cvs checkout' on a large
> tree will still kill the performance of your disk no matter
> which scheduler you are using (elevator, hybrid, the dumb one that
> we wrote as an example). The reasons, i suspect, are a mixture
> of the ones described above.
Interaction of the disk scheduler and softupdates? As I understand
the softupdate code, the disk scheduler would be ignorant of I/O that
depends on the completion of certain writes because issuing that I/O
to the lower layers is delayed until the equiv of bio_done occurs.
I don't see what use a disk scheduler would be unless it is possible
to push information about any partial ordering requirements down to
the level at which the scheduler can see and enforce them.
As you alude to, the scheduler cannot do anything useful unless it sees
a comprehensive picture of the pending I/O.
More information about the freebsd-geom