[PLUG] benchmarking webserver performance ?

[Russ Gilman-Hunt]

Thanks, to all of your responses.

The boss's old concept of "metric" is just the load, and then he uses
"top" to figure out what the problem could be. I do enjoy others :) And
I'm introducing him to the concept of histories and baselines.

I use mrtg-rrd for his servers, and cacti in another role at another
position.  What I'm looking for is a graph to be able to show him "yes,
response time is slowed down overall" or "this is a sudden drop in
response time and it corresponds to this other graph's spike here." 

He's not given me a budget, but then I haven't asked for one either. I'm
just using mrtg/mrtg-rrd to make some pretty graphs so he can see where
we stand.  For instance, the "storage server's" drive might have vast
tracts of land open, but it's increased 5 percent in usage over the past
two weeks, and is at 95 percent now.  You know, some sort of thing so he
can judge necessary resources.

So, really, I'm trying to figure out a few metrics to look at on his web
server to determine how slowly it's responding to requests.  

I suppose I could create a single base file and graph the download time
of that one file, as well as how long the server takes to acknowledge me
asking for it. I was thinking about using the index page of the main
site on the server; then they could see how changes to the index page
affects the performance.  If I use the single file, the size of the data
will always be the same.  Maybe what I should do is use a wget script
that downloads the index page and the requisites and just time the
length of time that script takes to process.

What I'm aiming at is something along the lines of "a customer places
the order within 4 minutes and his meal is delivered within 25" -- but
in webserver terms. I want to know when the meal is delivered.

Thanks again.
Russ