Bonnie Benchmark puzzle

[Gary Barnette]

For those that may not know, Bonnie is a great little disk benchmark tool
available on the net. Show Kbytes/second and %CPU time for different write,
read, and seek tests.

The setup:
I haven't run Bonnie for awhile on a box (HP RP8400) and then I run it
several times in a row. The first time it is run all of the write tests are
from 2-4 times as slow (depending on which test) as all the other subsequent
tests. The read tests are all very close, including the first time it is
run.

I'm sure it has something to do with buffer cache but I don't understand
what.

Anyone have an explanation? I would be glad to provide any other needed
information.

The reason I'm asking is that on another box of the same type, the write
test never does get better, even after the first running of Bonnie and this
box has a bad write performance problem. I am trying desperately to fix it.

Many Thanks,
Gary
[email]gbarnette@adelphia.net[/email]

[Mangled&Munged]

Hmmm.... I would have used Iozone. It can show
you the performance when the aggregate data set
fits in the buffer cache, and when it does not.

See: [url]http://www.iozone.org[/url]



"Gary Barnette" <gbarnette@adelphia.net> wrote in message
news:%ODsb.407$_i1.240424@news2.news.adelphia.net. ..

> For those that may not know, Bonnie is a great little disk benchmark tool
> available on the net. Show Kbytes/second and %CPU time for different

write,

> read, and seek tests.
>
> The setup:
> I haven't run Bonnie for awhile on a box (HP RP8400) and then I run it
> several times in a row. The first time it is run all of the write tests

are

> from 2-4 times as slow (depending on which test) as all the other

subsequent

> tests. The read tests are all very close, including the first time it is
> run.
>
> I'm sure it has something to do with buffer cache but I don't understand
> what.
>
> Anyone have an explanation? I would be glad to provide any other needed
> information.
>
> The reason I'm asking is that on another box of the same type, the write
> test never does get better, even after the first running of Bonnie and

this

> box has a bad write performance problem. I am trying desperately to fix

it.

>
> Many Thanks,
> Gary
> [email]gbarnette@adelphia.net[/email]
>
>