[Linux-cluster] stress-testing GFS ?

[Birger Wathne]

Michael Will wrote:

> iozone does test for a lot of different access patterns, and can 
> create nice spreadsheets including graphs
> from the point of view of a single node. It also has a multiple node 
> flag for running it across a cluster. See -+m and -t
> options. It knows how to use 'rsh' and can also be configured for any 
> other remote execution command by setting the
> enviroment variable RSH to say ssh or bpsh.
>
> Don't forget to post your benchmark results to this mailinglist ;-)
>
I used iozone and some homegrown scripts some years ago to test 
performance of various raid controllers as well as software raid on Sun 
systems. Always in single-node configurations.

The easiest way to communicate the performance of a raid controller to 
other people was a series of 3d surface plots. Sadly, OpenOffice doesn't 
have those, so I had to switch to that commercial office package. I 
tried gnuplot, but frankly.... compare the readability of the final plot 
with excel and there was no comparison :-/ Perhaps Matlab...

What I hoped for was something that also verified that the internal 
states of glm and the locking subsystem were as they should at every 
step of the test. Something that could certify that the hardware behaved 
as GFS expected it to when pushed more than test performance.

-- 
birger