Hi,

I would like to do some benchmark tests (SAN <=> local disks) on a 
solaris 10 server (V490)

Any idea how I would get reproducable results for read/write tests, or 
does someone know a good benchmark software/script ?

Thanks for your help
uxtuner

PS: my first idea was something like "time dd if=/dev/random bs=512
size=xxxxxx of=/dev/xxxx"

"uxtuner" <it@server-support.org> wrote in message news:f87v66$62i$1@svr7.m-online.net...
> Hi,
>
> I would like to do some benchmark tests (SAN <=> local disks) on a
> solaris 10 server (V490)
>
> Any idea how I would get reproducable results for read/write tests, or
> does someone know a good benchmark software/script ?
>

Have a look at filebench on solarisinternals.com

-- 
John.

uxtuner <it@server-support.org> wrote:
> I would like to do some benchmark tests (SAN <=> local disks) on a
> solaris 10 server (V490)

> Any idea how I would get reproducable results for read/write tests, or 
> does someone know a good benchmark software/script ?

Might look into iozone.org

rick jones
-- 
oxymoron n, commuter in a gas-guzzling luxury SUV with an American flag
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

In article <f87v66$62i$1@svr7.m-online.net>,
	uxtuner <it@server-support.org> writes:
> Hi,
> 
> I would like to do some benchmark tests (SAN <=> local disks) on a 
> solaris 10 server (V490)
> 
> Any idea how I would get reproducable results for read/write tests, or 
> does someone know a good benchmark software/script ?

The only benchmark which matters is your application,
so my first suggestion is to try and create some
reproducible representative tests using whatever you
will be using as your application. You can easily
check how reproducible your test is by running it
several times with the same conditions, but beware of
the effects of caching, which you may want to either
engineer in or out.

iozone is a good test to show you what works best on
a disk system, and you might be able to tune your
application to use the parameters iozone shows to be
particularly good (e.g. 64kbyte reads and writes are
significantly better than any other blocksize on some
of the filesystems I use, so I've made the application
use that size where I can). Using iozone to tell you
what performance you will get from your application
is probably very questionable though.

-- 
Andrew Gabriel
[email address is not usable -- followup in the newsgroup]