tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ken Anderson>
Subject Re: Which Apache-To-Tomcat Connector
Date Wed, 24 Apr 2002 17:53:22 GMT
I just did some similar benchmarking, though I didn't test with tomcat 
standalone, since I need Apache, and don't have standalone configured. 
I'm sure there are other factors, O/S, JDK, etc, but here's what I 
found, using mod_webapp, with Tomcat 4.04b2 on Sun JDK1.4 and Apache 
1.3.22 on RedHat 7.2.

Doing ./ab -n 1000 -c 100
Getting 15267 byte web page file; text and images (same file contents) 
in all three cases.

Apache only static content
- 52 req/sec avg. (no errors)

Apache through mod_webapp - getting same file - static content.
- 50 req/sec avg. (no errors)

Apache through mod_webapp - getting servlet/velocity generated page, 
based on sql result from mysql (simple select query through mm.mysql 
jdbc driver, using dbConnectionBroker pooling)
- 37 req/sec avg. (no errors)

I did notice that mod_webapp was a bit unstable; delivering some widely 
varying performance, and hanging 1 out of 20 tests I ran, though I have 
never seen this problem under normal loads. Also, since I didn't test 
standalone performance, I am not sure the problem I had was really with 

I tried hitting an WinXP dev box that runs TC standalone, but it 
couldn't handle the ./ab -n 1000 -c 100 load. It returned 'broken pipe' 
for anything over -c 5.

Ken Anderson

Cavan Morris wrote:
> I've recently done some benchmarking on mod_webapp and I wasn't too impressed.  I am
using it anyway though until I can compare to something better.
> The upshot was this.  Apache+tomcat was taking 5 times as long to serve a request as
tomcat standalone was.  Also, Apache+tomcat would choke on concurrent requests; sometimes
requiring an apache restart... Bad.
> It seams to me that Apache+tomcat should take on the order of the time it takes apache
to handle a request plus the time it takes tomcat to handle a request since that is basically
what happens.  Of course that's just my though without knowing any of the technical details
at all.  I could be an idiot.
> I'd be interested if anyone has performance data for the other connectors.  I used the
apache "ab" to get my data.  I compared my host through apache and my host standalone over
100 request and got under 20 requests per second with apache and over 100 requests / sec standalone.
> Cavan Morris

To unsubscribe:   <>
For additional commands: <>
Troubles with the list: <>

View raw message