Re: unwritten extents vs. fragmentation

["Rick Smith" <rgsmith72@xxxxxxxxxxx>]

    By preallocation are you referring to an ioctl after opening a file 
with XFS_IOC_RESVSP64? If so, does this only have an effect with 
unwritten=1?
    Does using the XFS_IOC_RESVSP64 ioctl guarantee contiguous extents when 
the file is written, or at least make more of an effort to ensure contiguous 
extent allocation? As you may remember from some of my previous emails, my 
goal is to have each file written as contiguous as possible and have each 
consecutive file as close to the previous file as possible. I have been 
experimenting with the realtime subvolume (due to the limited documentation 
speaking of reduced fragmentation), but I have not had much success so far. 
Thanks.
Rick

> From: Steve Lord <lord@xxxxxxx>
> To: Rick Smith <rgsmith72@xxxxxxxxxxx>
> CC: linux-xfs@xxxxxxxxxxx
> Subject: Re: unwritten extents vs. fragmentation
> Date: 07 Mar 2003 15:32:54 -0600
>
> On Fri, 2003-03-07 at 15:26, Rick Smith wrote:
> >      A couple questions. Does enabling unwritten extent flagging in XFS
> > ultimately reduce filesystem fragmentation? What is the downside of 
> enabling
> > unwritten extents other than slower write performance? Thanks.
>
> It should make no difference to physical fragmentation, and in the
> normal I/O path unwritten extents should not be used. They only come
> into play with space preallocation. In this case they will slow
> down I/O somewhat, but provide more security in that you cannot
> use unwritten extents to read old data. This is why the prealloc
> calls on -d unwritten=0 filesystems are restricted to root.
>
> Steve
>
> --
>
> Steve Lord                                      voice: +1-651-683-3511
> Principal Engineer, Filesystem Software         email: lord@xxxxxxx

_________________________________________________________________
The new MSN 8: smart spam protection and 2 months FREE*  
http://join.msn.com/?page=features/junkmail