On Wed, 19 Jan 2005 10:20:12 +0100, Mipam wrote:

> [SNIP]
> 
>> Does FFSv2 do journaling/logging?
> 
> FFSv{1,2} doesn't do journalling, if you wish so you can take a look at 
> LFS. FFS now uses soft updates that does ordened writes to make sure that 

How can any soft updates mechanism guarantee that e.g. the rename
operation, which is supposed to be atomic, will not leave both 
names (the old one and the new one) if it is interrupted by a crash in the
middle?

> the meta data is consistent. It performs better than a journalling fs (i
> wonder how many flames i'll recieve because of this statement :-)).
> There is a paper on this topic:
> 
> http://www.lcs.ece.cmu.edu/~soules/papers/seltzer.pdf

More precisely, it performs better than their own implementation of
journalling for FFS. Today's journalling filesystems have other
independent performance improvements over FFS, such as better
(tree-based) directory structure, which means that such journalling
filesystem could be much faster in certain situations that FFS+soft
updates.

Also, the time to recovery after crash is often more important than
performance - this thread is about large filesystems, so what will be the
fsck time for a 1TB FFS filesystem? I suspect it will be non-trivial.

I don't know if this is enough to qualify as a flame :-)

Bye	Pavel