Subject: Re: Google:Soc: FS
To: <>
From: Ignatios Souvatzis <>
List: tech-kern
Date: 05/11/2006 08:51:42
On Sun, May 07, 2006 at 03:14:45AM -0500, Sumantra Kundu wrote:
> Hi,
>   I was looking at the proposals for FFS and I have some pretty naive
> questions.
> - FFS Defragmentation algo: Could anyone please provide some pointers
> about the currently implemented schemes?  I just want to make sure
> that I am not duplicating efforts before I post my algo.
> - FS Caching: Is anything of that sort is available in the BSD
> community? What is the standad way to define an congestion event?

For both: are you aware of the classical work on this? FS caching (or 
rather, disk block caching as a service to the FS code) has been around
in original Unix (at least from V6/"Lion's Book" times) and, therefore
on BSD from the time Linux Torvalds was a toddler ;-)

Also, there should be some papers about the then "new" BSD filesystem
in the 4.3BSD book, or even earlier (was in the printed Ultrix or
the ConvexOS documentation in the end-eighties), explaining that
the BSD FFS variant does not do defragmentation, but rather _avoids_
excessive fragmentation by making sure big files are split, so there is
a good chance that you can extend another file near its original 
block allocation.

As for the NetBSD Unified Buffer cache, there's a Usenix paper about
that by Chuck Silvers:

seal your e-mail: