[OpenAFS-devel] Patch to implement default tuning proposal discussed a while ago

[Troy Benjegerdes]

On Fri, Aug 12, 2005 at 01:14:06PM -0400, chas williams - CONTRACTOR wrote:
> In message <20050812161014.658FC1BAA9@citi.umich.edu>,Jim Rees writes:
> >And why did you change the assumed average file size to 32KB?  This might
> >actually be a good number but I wonder if you have data to back it up.  I
> >just measured my own client cache and came up with 30.2KB so you could be
> 
> this is curious.  my avg file size is 38k.  so this is an interesting
> choice.  however, wouldnt a better choice be the median file size?
> 75% of my files are 10k or less.  just a thought.  everyone loves
> statistics.

How do you measure size of files in the cache? What do we lose by having
the average file size being too big?

I just analyzed the couple of volumes I'm using for afsroot, which
I'm most interested in having good default tuning parameters for, and in
the entire volume, I get 75% or so of the files less than 10k.

What's more interesting than probably the median or average file size is
the distribution of file sizes. (although maybe median captures some of 
what I'm interested in).

So, I did the following:
find /some/afs/path -noleaf -type f -printf '%s\n' | sort -n > filesize

gnuplot
gnuplot> set logscale y
gnuplot> plot 'filesize'

http://scl.ameslab.gov/~troy/debian-nodeimg-size.png
http://scl.ameslab.gov/~troy/afsserver-vicepd.png

The second graph is looking at the /vicepd partition of one of my AFS
servers directly.

-- 
--------------------------------------------------------------------------
Troy Benjegerdes                'da hozer'                hozer@hozed.org  

Somone asked me why I work on this free (http://www.fsf.org/philosophy/)
software stuff and not get a real job. Charles Shultz had the best answer:

"Why do musicians compose symphonies and poets write poems? They do it
because life wouldn't have any meaning for them if they didn't. That's why
I draw cartoons. It's my life." -- Charles Shultz