Yes, I understand this, and what vfs.lfs.pagetrip is (I can run sysctl
-d ;).  The thing is, even higher values than that--even by a few
thousand pages, dramatically decrease throughput.  I figured it would
be as simple as your account sounds, but it isn't.

I have found the best overall performance in real-world tasks comes
with using the aggregate bandwidth of all disks in my ccd
(vfs.lfs.pagetrip=8596), using an LFS segment size figured from the
same (17199 blocks), and a stripe size of 5733 blocks (1/3 segment
size).

--Blair

On 12/11/06, Thor Lancelot Simon <tls@rek.tjls.com> wrote:
> On Mon, Dec 11, 2006 at 12:15:16AM -0500, Blair Sadewitz wrote:
> >
> > Now, this is where it gets strange.  To arrive at a value for
> > vfs.lfs.pagetrip, I was told to divide the bandwidth of my disk (in
> > the case of the ccd, it is ~162000000 by 4096 and then divide that by
> > four, producing a result that should be bigger than your chosen
> > segment size.  However, if I simply divide the bandwidth by 4096 (size
> > of one page) and set vfs.lfs.pagetrip to, for example, 39950, bonnie++
> > results tend toward one third better write performance, and other
> > improvements all around.  If anyone has the hardware to create and
> > benchmark LFS filesystems to investigate this, espically on raidframe
> > or ccd.
>
> You underestand that bandwidth is not the only important measurement of
> disk performance, right?  If you set the pagetrip value so that writes
> are started only when a full second of data is available to be written,
> you're going to significantly increase the _latency_ of the filesystem.
>
> Where you set the balance point best will depend on your application but
> for most applications except for video writing (only, with no reads) the
> value you're suggesting is too high.
>
> --
> Thor Lancelot Simon	                                     tls@rek.tjls.com
>   "The liberties...lose much of their value whenever those who have greater
>    private means are permitted to use their advantages to control the course
>    of public debate."					-John Rawls
>


-- 
Support WFMU-FM: free-form radio for the masses!

<http://www.wfmu.org/>
91.1 FM Jersey City, NJ
90.1 FM Mt. Hope, NY

"The Reggae Schoolroom":
<http://www.wfmu.org/playlists/RS/>