Flood output

[Ken Youens-Clark <ky...@gmail.com>]

Hi,

I just discovered flood, and I'm interested in writing my own report  
based on the raw output.  The only documentation I can find is here:

	http://cvs.apache.org/viewcvs.cgi/httpd-test/flood/docs/docbook/ 
flood.xml?rev=1.8&view=markup

Last updated on April 15, 2004, where it says at the very end:

     <!-- This section needs to be written. Volunteers? [jacekp] -->
     <chapter>
         <title>Analyzing Flood output</title>
         <para>No information available at this time.</para>
     </chapter>

I've done a bit of reverse engineering with the  
"examples/analyze-relative" script and believe that the columns  
represent the following:

1  ???
2  connect
3  write
4  read
5  close
6  response
7  ???
8  URL

As you can see, I can't figure out the 1st and 7th columns.  I would  
love it if someone could help me out on this.  I'm working on a Perl  
version of the "analyze-relative" script so I can expand what it does.

Perhaps a question I should get answered before all of the above is  
whether flood is still being developed, if it is reliable/doing what I  
think, etc.?  Should I be using flood or can someone point me to a  
better program for stress-testing a web server with a bunch of URLs.

ky

Re: Flood output

[Jacek Prucia <ja...@acn.waw.pl>]

On Thu, 6 Jan 2005 08:34:55 -0600
Ken Youens-Clark <ky...@gmail.com> wrote:

> Hi,
> 
> I just discovered flood, and I'm interested in writing my own report  
> based on the raw output.  The only documentation I can find is here:
> 
> 	http://cvs.apache.org/viewcvs.cgi/httpd-test/flood/docs/docbook/ 
> flood.xml?rev=1.8&view=markup
> 
> Last updated on April 15, 2004, where it says at the very end:
> 
>      <!-- This section needs to be written. Volunteers? [jacekp] -->
>      <chapter>
>          <title>Analyzing Flood output</title>
>          <para>No information available at this time.</para>
>      </chapter>

I've left this section empty, because at the time i wrote this, other topics
seemed much more important.

You can get some more information on flood output here:

http://www.clove.org/flood-presentation/flood.pdf (page 77)
 
> I've done a bit of reverse engineering with the  
> "examples/analyze-relative" script and believe that the columns  
> represent the following:
> 
> 1  ???
> 2  connect
> 3  write
> 4  read
> 5  close
> 6  response
> 7  ???
> 8  URL
> 
> As you can see, I can't figure out the 1st and 7th columns.  I would  
> love it if someone could help me out on this.

1 Start
7 ID

Start column marks beginning of a request, while ID identifies client (farmer
process/thread).
 
> Perhaps a question I should get answered before all of the above is  
> whether flood is still being developed,

Well... sort of yes and no. Personally I haven't lost interest in flood, but
due to very specific circumstances, i'm not able to work on it (yet). Since
all other flood developers seem to be also busy, several people figured out,
that this project is dead (*), which isn't really true.

> if it is reliable/doing what I think, etc.?

Well, I've used flood in real life scenarios, and i must say flood is quite
efficient. It might be missing few features, but is way faster than other
stress testing packages.

> Should I be using flood or can someone point me to a  
> better program for stress-testing a web server with a bunch of URLs.

I would suggest to stick with flood. Drop us a line every time you feel flood
is missing some special feature. We might not be able to implement it
instantly, but it will probably end up in our TODO list.

(*) for example see this link:
http://www.apache.org/~rbowen/presentations/oscon2004/apacheperformance/slide13.html

regards,
Jacek Prucia