On 8/5/06, Rob Kirkbride <rob.kirkbride@xxxxxxxxx> wrote:
>
>
> On 05/08/06, Joshua Slive <joshua@xxxxxxxx> wrote:
> > On 8/5/06, Rob Kirkbride <rob.kirkbride@xxxxxxxxx> wrote:
> > > On 8/4/06, Joshua Slive <joshua@xxxxxxxx > wrote:
> > >
> > > >
> > > On 8/4/06, Rob Kirkbride <rob.kirkbride@xxxxxxxxx> wrote:
> > > > Hi all,
> > > >
> > > > I need to understand exactly what this figure means.
> > > > Does the figure include any time for the response to be sent back to
> the
> > > > client, or is it the time that the web server has completed the
> > > processing?
> > > > We're getting some long duration and I need to understand if we've
> got
> > > > network problems at our end, or if it may be because the user is
> running
> > > > over a slow connection.
> > > >
> > >
> > > It is the time from the request is received to the time apache httpd
> > > completely finishes with the request. That will include the time
> > > required to send the file over the network.
> > >
> > > Joshua.
> > >
> > > Thanks Joshua.
> > > So the upshot is that it's a useless statistic to see how loaded/busy
> the
> > > website is then? We have to achieve a certain percentage within a
> certain
> > > time. Is there anything else I can do?
> >
> > A certain percentage of what within a certain time? How busy a
> > website is depends on the performance of the network and the clients
> > as well as the server, so your question doesn't seem well-defined.
> >
> > Joshua.
> >
> >
> Sorry you're right. We basically have bands whereby if we achieve >=99%
> within 1 second the client pays us. If we're between say 98 and 99% the cost
> is neutral, less than 98 and we pay them.
> It's slightly more complicated than that but that's the basic idea.
> We have to ensure we have enough web-servers to deal with the requests and
> this is there way of checking.
I'm not an expert in this stuff.
If you are concerned about the time to generate complicated dynamic
content, then you can build a timer into the application itself. For
example, this is what google does to tell you on each search response
how long the response took to generate.
Otherwise, you could consider using a monitoring application on the
local network to periodically check performance of your site. This
would eliminate the client and network as possible variables, but
would only provide a statistical sample rather than a definitive
measure of all response times.
I don't think there is any way for apache httpd to answer this question.
Joshua.
|