In order to measure reliably some worst-case latencies, is there a way
to have the system (cleanly) drop as much as possible of its
page/directory cache? Being able to specify which device would be a
plus ;-)
OG.
On Tue, 31 October 2006 18:12:04 +0100, Olivier Galibert wrote:
>
> In order to measure reliably some worst-case latencies, is there a way
> to have the system (cleanly) drop as much as possible of its
> page/directory cache? Being able to specify which device would be a
> plus ;-)
grep -A5 drop_caches Documentation/filesystems/proc.txt
J?rn
--
Joern's library part 9:
http://www.scl.ameslab.gov/Publications/Gus/TwelveWays.html
On Tue, Oct 31, 2006 at 06:23:10PM +0100, J?rn Engel wrote:
> On Tue, 31 October 2006 18:12:04 +0100, Olivier Galibert wrote:
> >
> > In order to measure reliably some worst-case latencies, is there a way
> > to have the system (cleanly) drop as much as possible of its
> > page/directory cache? Being able to specify which device would be a
> > plus ;-)
>
> grep -A5 drop_caches Documentation/filesystems/proc.txt
Excellent, thanks.
OG.