Thanks, Daniel. What I really need is some way to measure how badly loaded a
web server is, and I think what I'm calling "page render time" may be about as
good as I can get.

I'm hoping that by subtracting time_pretransfer from time_starttransfer, I get a
time measurement that doesn't include network holdups, and is all therefore due
to the web server being hammered (or not).

Perhaps instead of calling it "page render time" I should have called it
something different, such as "server latency" or "server being beaten to death
time" or something... :)

The particular servers I need to watch this for are running off to a mainframe
database for information, so the "render time" will include dragging stuff of
its own disk (or cache), querying the mainframe, shaping the results and shoving
them back down the pipe.

The people that own the servers don't even have a measurement for deciding how
they're performing. They just start calling us when "the server seems to be
struggling"... It'll be really funny if our support folks can tell them "well,
the page test script is averaging 1.5 seconds, of which 0.3 seconds is taken by
the server and the rest is due to the local network. If you're seeing bigger
delays, maybe your isp is 'struggling'..."

Ralph

Daniel Stenberg wrote:

> On Tue, 10 Dec 2002, Ralph Mitchell wrote:
>
> > Am I right in thinking that I can calculate the page rendering time from:
> >
> > time_starttransfer - time_pretransfer
>
> I don't think it necessarily is only the page rendering time. But it surely
> is the time from after the connection is established until the first byte has
> been transfered.
>
> > If not, what would the above give me?
>
> If we ignore the transfer time (for the first byte to get across to you), it
> is probably also the delay in the web server from when it gets the first
> pieces of request until it finds the HTML or runs the script or whatever it
> does.
>
> Also, most scripts that render a page will start sending out HTML to the
> client before the full output is done, so the transfer starts before the full
> page is "rendered".
>
> Estimating page rendering times really must be next to impossible on the
> client's side.
>
> These are all my weird assumptions. Feel free to fill in your own! ;-)
>
> --
> Daniel Stenberg -- curl, cURL, Curl, CURL. Groks URLs.
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf

-------------------------------------------------------
This sf.net email is sponsored by:
With Great Power, Comes Great Responsibility
Learn to use your power at OSDN's High Performance Computing Channel
http://hpc.devchannel.org/
Received on 2002-12-11