On Thu, Mar 25, 1999 at 01:40:00PM -0500, Mason Loring Bliss wrote:
> On Thu, Mar 25, 1999 at 12:30:51PM -0500, Jim Wise wrote:
> 
> > _The Designn and Implementation of the 4.4BSD Operating System_, by
> > McKusick et al has a section describing LFS.  It's not very long, but it
> > gives a good overview/introduction.
> 
> Hm... Things are busy at work, so I haven't yet looked at the online
> references folks have sent, but I do happen to have my copy of the 4.4BSD
> book with me. (I'm reading it through straight, and hadn't reached the bit
> about LFS yet.)
> 
> >From my reading, it actually seems that LFS is *less* stable in some ways
> than FFS, not more... My interest in soft updates / JFS is the stability
> you see when, for instance, your power is dropping during the middle of a
> write. LFS doesn't seem to really be aiming at the same thing as JFS or
> soft updates. I don't see it as being something I can point to as being
> the right thing for the place at which I'm working at the moment, whereas
> soft updates or a journalled filesystem would be a big win.

I don't understand what you mean by "stability".  FFS with soft updates is
exactly like normal FFS in terms of failure recovery, it just reduces
synchronous write overhead while the filesystem is running.

LFS will at least let you boot your system faster, at the expense of
predictably losing the last (checkpoint interval) worth of updates
(whereas in FFS you may lose _some unpredictable fraction of them_), by
rolling back to the last checkpoint in the log.  It _also_ eliminates
the synchronous write overhead that soft updates do in FFS; in fact,
it may do a better job of it.