On Tue, 28 Nov 2000, Larry Leszczynski wrote:
> Hi All -
>
> > I'm hoping for some enlightenment about how KeepAlive is implemented in
> > Apache and whether KeepAlive even comes into play when front-end and
> > back-end mod_perl servers communicate with each other via HTTP.
>
> I suppose I shouldn't have started off my previous post with such a
> general-sounding comment. Thanks for all the pointers to the KeepAlive
> section in the mod_perl guide, which I have read before and which doesn't
> answer the questions I was asking below.
>
> I'll try rephrasing, I'm still hoping for info regarding the questions
> that follow:
>
> mod_backhand takes advantage of KeepAlive to speed up communications
> between a front-end server and a set a back-end servers that feed data to
> the front-end. I'm trying to figure out how that works and if I can take
> advantage of KeepAlive the same way using mod_perl front-end and back-end
> servers but without running mod_backhand.
>
> Suppose front-end server A is handling user requests. In the process of
> handling a front-end request, suppose I use LWP or equivalent to make a
> HTTP request from A to a back-end server B to get some data that is
> needed. Assuming all the right headers are set for KeepAlive to work
> (content length, etc.), can the connection between A and B even take
> advantage of KeepAlive for the next time A makes the same request to B?
>
> One problem is that I'm not sure what processes would actually be
> "keeping" the ends of the "kept alive" connections. At each end, would it
> be the parent httpd process, or the individual httpd child process that
> made/answered the request?
>
> I'm thinking that if A had to fork a CGI that in turn talked to B, the
> kept-alive connection would be lost as soon as the CGI process on A died
> (socket timeouts notwithstanding). But what if the request from A to B is
> made from within a mod_perl module, or within an Apache::Registry script?
>
> Along the same line of thought (assuming this has made any sense so far),
> what happens when you throw ProxyPass/ProxyPassReverse into the mix? What
> (if anything) can be done to take advantage of KeepAlive then?
Before trying to answer your question, let me ask you another
question. What's the average run time of you mod_perl scripts? Is it 2 to
5 msecs? Or is it 0.5 to 1 sec? If you are in the former range this
question makes sense, well may be. If you are in the latter range, you are
wasting your time.
KeepAlive comes to overcome the overhead of initiating connections between
the browser and the server. It's useful when you serve static pages,
because they are delivered in matter of a few milli-seconds. Which is
somewhat comparable with the overhead of creating the connection... or
very fast mod_perl code whose run time comparable to delivery of static
pages (we have simple ads serving at 4-6msec).
KeepAlive helps a lot with SSL connections (https) where the handshake is
a few (4-5?) times slower compared to http, but it's still significant in
the same range of run times.
If your average run time is 10-100 slower than 5 msec, this overhead is
insignificant since it comes down to something like 0.01% of the total
response time over http and probably 0.05% over https.
Now do you still want the KeepAlive?
_____________________________________________________________________
Stas Bekman JAm_pH -- Just Another mod_perl Hacker
http://stason.org/ mod_perl Guide http://perl.apache.org/guide
mailto:[EMAIL PROTECTED] http://apachetoday.com http://jazzvalley.com
http://singlesheaven.com http://perl.apache.org http://perlmonth.com/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
