Subject: Re: Journaling for FFS
To: Bill Studenmund <firstname.lastname@example.org>
From: Manuel Bouyer <email@example.com>
Date: 10/02/2006 22:10:26
On Mon, Oct 02, 2006 at 12:56:25PM -0700, Bill Studenmund wrote:
> On Mon, Oct 02, 2006 at 09:16:29PM +0200, Manuel Bouyer wrote:
> > On Mon, Oct 02, 2006 at 09:09:49PM +0200, Jochen Kunz wrote:
> > > On Mon, 2 Oct 2006 09:26:04 -0700
> > > Bill Studenmund <firstname.lastname@example.org> wrote:
> > >
> > > > If someone wants to work on journaling for ffs, please don't
> > > > discourage him or her.
> > > >
> > > > LFS and FFS work well for different work loads. There are workloads
> > > > for which either one of them will blow the other away. As such,
> > > > neither one will replace the other for performance use under certain
> > > > workloads; we need both.
> > > This is a good point. I didn't think of this. The same for Jorgen
> > > Lundman's remark about adding jornaling to FFS like a mount option just
> > > like softdeps.
> > >
> > > Matt: Go for it! :-)
> > > (One thing to take into consideration: It may be more valuable to have
> > > jornaling for FFSv2 first, then for FFSv1. A FFSv2 can be much larger
> > > then a FFSv1...)
> > I'm not sure it makes a big difference supporing only one, or both.
> Depends on how you do it. If you go for a physical block journal, it's the
> same for FFS1 and FFS2. A logical journal (which I dislike, but wouldn't
> block if it showed up as a complete thing) will be different. Well,
> journal replay will care.
Yes, but even in this case it shouldn't be a big issue.
Probably less than adding quota support to FFF2 :)
Manuel Bouyer, LIP6, Universite Paris VI. Manuel.Bouyer@lip6.fr
NetBSD: 26 ans d'experience feront toujours la difference