You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@tomcat.apache.org by Jakob Hummes <hu...@castify.net> on 2000/08/02 21:40:47 UTC
Tomcat Servlet Performance Issues
Hi all,
I've found some other messages in the archives, which tend to support my
measurements.
I've a servlet, which does heavy communication with the DB and an own
server. No JSP at all. Created pages are very, very small (less than 1
kByte).
Testing this servlet with Tomcat 3.1. standalone on a Windows2000 500
MHz, 512 MB Ram, SUN Java 1.2.2 with JIT, no atumatic reload, performing
concurrent request from 5 other machines, I got a lousy performance of 8
pages per second.
The same test with the "Hello World" example got me 11 pages per second.
The I setup IIS to act as the Web server rediercting requests to the
test servlets. It was a bit faster (around 10%), but no great
difference.
Right now, it's OK, but very soon I have the requirements of serving at
least 50 pages per second from one machine. (Yes, I know that Tomcat
scales that means that I can add hardware; but hardware is also a cost
factor and I cannot tell clients that they need 20 machines to serve 200
pages a second).
Is there any way to speed up the performance per machine?
Thanks,
- Jakob
Re: Tomcat Servlet Performance Issues
Posted by Steve Ruby <st...@rubysolutions.com>.
Jakob Hummes wrote:
>
> Hi all,
>
> I've found some other messages in the archives, which tend to support my
> measurements.
>
> I've a servlet, which does heavy communication with the DB and an own
> server. No JSP at all. Created pages are very, very small (less than 1
> kByte).
>
> Testing this servlet with Tomcat 3.1. standalone on a Windows2000 500
> MHz, 512 MB Ram, SUN Java 1.2.2 with JIT, no atumatic reload, performing
> concurrent request from 5 other machines, I got a lousy performance of 8
> pages per second.
>
<snip>
> Is there any way to speed up the performance per machine?
>
> Thanks,
> - Jakob
Have you evaluated any of the newer tomcats... I'm running 3.2b2 right
now
which has been stable for me, and although I have no numbers it feels
faster
than 3.1 did.
since you have the testing environ. setup maybe you could plop a newer
tomcat
in there easily.
How to restrict access to 8007 (Was Re: Tomcat Security Issues)
Posted by Holger Klawitter <ho...@klawitter.de>.
Thomas Stahl wrote:
>
> Hi all,
>
> I use Tomcat with MS IIS 4.x. and I want to ensure access to
> Tomcat only through my MS IIS instance. So I deleted the Connector
> definition in server.xml to prevent direct http access via port 8080.
> But where can I configure the Ajp12ConnectionHandler(port 8007) only
> to acceppt requests from my IIS on a specific host.
> Any ideas?
I tried the "host" attribute, but it didn't work :-(
Regards,
Mit freundlichem Gruß,
Holger Klawitter
--
Holger Klawitter +49 (0)251 484 0637
holger@klawitter.de http://www.klawitter.de/
Tomcat Security Issues
Posted by Thomas Stahl <th...@debis.com>.
Hi all,
I use Tomcat with MS IIS 4.x. and I want to ensure access to
Tomcat only through my MS IIS instance. So I deleted the Connector
definition in server.xml to prevent direct http access via port 8080.
But where can I configure the Ajp12ConnectionHandler(port 8007) only
to acceppt requests from my IIS on a specific host.
Any ideas?
Thomas Stahl
Re: Tomcat Servlet Performance Issues
Posted by Jakob Hummes <hu...@castify.net>.
James,
thanks a lot for this link. I'll try to recreate the testbed from
Caucho for Tomcat and will post the data as soon as I have them.
Probably not before the end of the next week.
The thing what hit me most was the performance of a very simple page by
the Webserver (IIS or Apache), which gives one the upper limit. Last
time I ran the test I got less than 200 pages by either Web server;
that's a far cry from the published benchmark at Caucho and my hardware
should even perform better.
So, I'll go after this next week at keep you posted.
Thanks to all for the answers (with the exception of the guy who assumed
that I'm a bloody Java/JDBC beginner;-),
- Jakob
James Cook wrote:
>
> I would assume that something is *terribly* wrong with the testbed. Check
> out the benchmark that caucho put together at
> http://www.caucho.com/articles/benchmark.xtp
Re: Tomcat Servlet Performance Issues
Posted by James Cook <ji...@iname.com>.
I would assume that something is *terribly* wrong with the testbed. Check
out the benchmark that caucho put together at
http://www.caucho.com/articles/benchmark.xtp
It has the HelloWorld servlet running around 83 pages/sec on a P2/266 64MB
ram. Not sure how many clients that test represents. Tomcat performance is
still way, way behind the 500+ pages/sec achieved with Resin.
The server was running on IBM's JDK, the OS was Linux. Linux hasn't been
proven to be *faster* at web serving than NT. Just serving the page (Sorry
Linux guys).
You may also want to look at the benchmarks posted by orion at
http://www.orionserver.com/benchmarks/benchmark.html
jim
----- Original Message -----
From: "Jakob Hummes" <hu...@castify.net>
To: <to...@jakarta.apache.org>
Sent: Wednesday, August 02, 2000 3:40 PM
Subject: Tomcat Servlet Performance Issues
> Hi all,
>
> I've found some other messages in the archives, which tend to support my
> measurements.
>
> I've a servlet, which does heavy communication with the DB and an own
> server. No JSP at all. Created pages are very, very small (less than 1
> kByte).
>
> Testing this servlet with Tomcat 3.1. standalone on a Windows2000 500
> MHz, 512 MB Ram, SUN Java 1.2.2 with JIT, no atumatic reload, performing
> concurrent request from 5 other machines, I got a lousy performance of 8
> pages per second.
>
> The same test with the "Hello World" example got me 11 pages per second.
>
> The I setup IIS to act as the Web server rediercting requests to the
> test servlets. It was a bit faster (around 10%), but no great
> difference.
>
> Right now, it's OK, but very soon I have the requirements of serving at
> least 50 pages per second from one machine. (Yes, I know that Tomcat
> scales that means that I can add hardware; but hardware is also a cost
> factor and I cannot tell clients that they need 20 machines to serve 200
> pages a second).
>
> Is there any way to speed up the performance per machine?
>
> Thanks,
> - Jakob
Re: Tomcat Servlet Performance Issues
Posted by Bernhard Wraase <bw...@netkosmos.de>.
> Is there any way to speed up the performance per machine?
My test environment was quite different from yours but may give you a hint.
Apache/Jserv was the fastest.
I use just servlets with a jdbc-solid conection and some of the servlets
build
a ftp connection. (5 different servlets)
As a client tool I take jmeter.
In this configuration is Apache/jserv much more faster then Apache/tomcat
and Apache/Resin. The rerformance of both are similar, sometimes is one
better than the other. But the differences are about 10 percent.
I' ve just tested it on Linux because our software must run on Linux.(You
know
the client...) I took jserv 1.1.1
Bernhard