Mailing-List: contact new-httpd-help@apache.org; run by ezmlm
Precedence: bulk
Reply-To: new-httpd@apache.org
Date: Thu, 16 Aug 2001 23:15:30 -0700
From: Brian Pane <bpane@pacbell.net>
Subject: Re: Performance numbers..... ;(
To: new-httpd@apache.org
Message-id: <3B7CB682.5070308@pacbell.net>
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii; format=flowed
Content-transfer-encoding: 7BIT
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.3) Gecko/20010801
References: <15228.16811.269589.434751@critterling.garfield.home>
Status: O

Victor J. Orlikowski wrote:

>Hi all,
>
>Was running some performance tests on AIX...
>And oddities popped up between 1.3 and 2.0 (latest CVS of both).
>
>Requests/sec.
>                    no keepalive          keepalive
>2.0 - prefork          420                  590
>2.0 - threaded         390                  580
>1.3                    420                  700
>
>(Semi) Good news: Without keepalives, prefork performance in 2.0
>matches 1.3. Threaded performance comes close.
>
>Bad news: 1.3 trashed 2.0 on keepalives. Especially odd is the fact
>that the threaded and prefork numbers here are *so* close.
>
>So, threaded isn't getting trounced too badly, currently. But
>keepalive performance in 2.0 has me worried.
>
>Anyone have any ideas?
>
As a first step, do you have a means of capturing a
system call trace (with something like strace)?  If
there are any obvious differences in the system call
profile between 1.3 with keepalives and 2.0-prefork
with keepalives, that may help to isolate the problem.

If the system call behavior isn't substantially
different, I'd expect to see a higher percentage
of usr (as compared to sys) CPU time in 2.0 with
keepalives.

My hypothesis is that the extra brigade processing
involved in keepalive handling in 2.0 may be contributing
to the slowness.  But it's tough to tell without
fine-grained profiling tools.  Can anybody with
a statement-level or basic-block-level profiler
(e.g., Quantify) post an analysis of the bottlenecks
in 2.0?

--Brian