[httperf] httperf benchmarking and reporting

Theodore Bullock tbullock at canada.com
Fri Jul 18 22:42:49 PDT 2008


On Fri, 2008-07-18 at 14:51 -0600, Theodore Bullock wrote:
> On Fri, 2008-07-18 at 11:32 -0700, Mark Nottingham wrote:
> > Would it be worth having a mode where httperf reports its results in
> > a  
> > more normal and machine-friendly way, so that tools can depend on  
> > output, and we have more freedom to change the output intended for  
> > people?
> 
> At WWW2007 last year I had a discussion with a couple people about this
> very topic.  Here is a bit of what I came up with.
> 
> CSV
> 
> A flat text file with the data entered in a row.
> 
> Would need a new option to specify the file name.  If the file doesn't
> exist we create it, with a "header" row.  If it does exist, we just
> append our data to the end of it.
> 
> Thus each consecutive run of httperf grows the data set.
> 
> I think this is the most appropriate solution for global "end of
> experiment results".
> 
> It would also be nice to collect a histogram of the measures from each
> connection/session.  Thoughts on this?
> 
> -Ted
> 
> 
Haha,

I didn't finish my thought here.  That's what you get when you are in a
panic trying get in just one last email....

Anyways, some of the web guys at the conference suggested RDF as a good
way of integrating some sort of semantic meaning to the data.

Also suggest was a generic XML file of some sort.

I'm not certain the advantage of this over csv in our case though.
Certainly the semantic nature of RDF could be useful, however, who
cares....

That yaml stuff also seems interesting, however I don't understand what
it provides that a csv file does not.  For instance, how do I go from a
yaml file to a chart?

Are there any other options out there?

-- 
Ted Bullock <tbullock at canada.com>
http://www.comlore.com
High Voltage Software Engineer



More information about the httperf mailing list