[Samba] Standardized Benchmarking?

AragonX aragonx at dcsnow.com
Tue May 24 13:34:17 GMT 2005


<quote who="Peter Szmrecsanyi">
> Good idea! Bonnie++ is good, I also like tiobench as it benchmarks the
> block
> device's IO capabilities...
>
> I used a stopwatch as well, I suppose a small program that will copy a
> large
> file and give back a rate would be useful in benchmarking.
>
> I suppose that as far as standardizing, well all you need to is to specify
> the file size and type (gz/zip for example). A good idea would be to bench
> small files, reasonably sized files yet smaller than RAM and then a big
> file
> (say 4GB) which must be bigger than RAM... That way you can get an idea of
> what RAM -> RAM and Disk -> Disk transfers are like and how they differ,
> with a medium sized file you'd get a mixture of RAM and Disk.

I agree that different size files should probably be tested.  If some of
the Samba developers want to chime in, that would be great (hint hint). 
I'm thinking one large transfer and one small.  I currently use a 4gb file
for the large transfer and have used 10,000 100k files in about 100
different directories.

I looked through my 30+ Compaqs I have sitting in my back room.  I didn't
see any of the PPros left.  I must have either gave them away or used them
somewhere else.  Let me continue checking.  I know I should have 2
somewhere.

If you could, please post your results on the site.



More information about the samba mailing list