Michael Lampe wrote:
Is this all I need (I already have applied the ide write barrier patch)
to safely use disk write back caching?
(which IDE write barrier patch is that, out of curiousity?)
The one that SuSE ships with their latest kernel. Due to Jens Axboe, I
think. (Attached.)
This stuff is untested by anyone here at SGI, so YMMV. It is
also not going to work in the presence of unwritten extents
(which were implemented after this change), some additional
code would be needed there.
Pity, that. Beeing forced to disable write caching is major performance
killer.
-Michael
I think write back caching is a lame IDE performance hack. SCSI disks
use TCQ to achieve write performance without compromising data
integrity. (XFS guys - Linux does this, right?) I wonder if it would
be more productive to pursue the IDE-TCQ stuff that's coming down the pipe.
Regards,
Jeremy Jackson
|