Re: state or future of LFS?

[Andrew Doran <ad%netbsd.org@localhost>]

On Fri, Apr 10, 2009 at 02:18:00PM +0200, Niels Dettenbach wrote:

> what is the current usability state of the LFS (Log-structured File System)? 

It is broken.

> Is it ready for productional use or still to understand as under development? 
> Is there any active roadmap? 
> 
> Did someone have practical experiences (stability, performance, handling...) 
> with LFS within productive environments? Are there any other interesting 
> alternatives under NetBSD? 

The unfortunate truth is that BSD-LFS started out as research project and
has not yet progressed beyond that stage, despite the best efforts of many
developers and advocates. For NetBSD it was also somewhat interesting for
hobby and toy applications.

> Did someone played or worked with other filesystem alternatives for large
> hard disks (i.e. some cluster FS's a.o.) then the default FFS with
> interesting results?
> 
> I'm using NetBSD 4.0 and 4.99.73 (and sometimes other *BSD's) for different 
> productive internet servers (i.e. email servers) - under xen and wo - on SAS 
> based disk RAID's (10 and 5)  and my idea is to use / switch over to a LSF 
> for my larger email-storages. 
> 
> My main target's are:
>  - fast recovery (i.e. after a crash / power disruption)
>  - fast file access (i.e. mail space under cyrus imap)
>  - minimized data losses after crash or medium failures

5.0's UFS2+logging meets the above targets: multi-TB file systems, no fsck,
near async performance and safe in the presence of hardware caching.