Re: WAPBL vs. lfs?

[Matthias Scheler <tron%zhadum.org.uk@localhost>]

On 23 Aug 2008, at 0:16, Toru Nishimura wrote:
> The performace degregration of FFS WAPBL journaling is quite  
> noticiable
> for slow computers.

That might be true. I only tested it on a Xeon 3040 system.

>  Journaling is a sort of "high availablity" feature in
> return of cost, for CPU and disk IO handwidth.

WAPBL actually saves IO bandwidth, see below.

> I would pick FFS softdep rather than log which should be carefully
> intended for application domain where it does matter.

And that would probably be the wrong choice. I've benchmarked soft
dependences and WAPL on the same hardware. The result is that WAPBL
is faster. You can find my results here:

http://zhadum.org.uk/2008/08/23/wapbl-vs-soft-dependences/

        Kind regards

--
Matthias Scheler                           http://zhadum.org.uk/