Hi Jerry,
I think you make a good recommendation, explaining the issue of write back via cache vs.
immediately hitting the disk. I also think your mail explains why we should not change the
default. Instead, lets change the documentation, to explain that those not needing
consistency can use this performance optimization.
Jon.
--
Sent from my iPad
On Jan 11, 2013, at 10:42, Jerry James <loganjerry(a)gmail.com> wrote:
I've been using the vexpress image with QEMU to try to track down
some
ARM-specific problems in two of my packages. The VM was unbearably
slow at first (by which I mean that a simple "yum upgrade" took over 6
hours), until I changed:
-sd "$IMAGE"
to:
-drive if=sd,cache=writeback,file="$IMAGE"
in boot-vexpress and boot-vexpress+x. I'm well aware of the danger of
using writeback, but ... it's a VM. With the script as shipped, my
hard drive light was on constantly, and the entire host was slowed
down dramatically due to insufficient disk bandwidth. With writeback
on, I can re-create the VM and reinstall the packages of interest in
very little time, certainly less than even simple operations were
taking with write-through on. Is there some reason why changing the
scripts to use writeback would not be a good idea?
Regards,
--
Jerry James
http://www.jamezone.org/
_______________________________________________
arm mailing list
arm(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/arm