Mailing-List: contact tomcat-user-help@jakarta.apache.org; run by ezmlm
Message-ID: <2B26E094BB13D3118FB3006008214FA5F06F7F@vdlaexc0.orange.ch>
From: Kitching Simon <Simon.Kitching@orange.ch>
To: "'tomcat-user@jakarta.apache.org'" <tomcat-user@jakarta.apache.org>
Subject: RE: Performance
Date: Wed, 4 Oct 2000 15:48:19 +0200 
MIME-Version: 1.0
Content-Type: text/plain


> -----Original Message-----
> From:	Lacerda, Wellington (AFIS) [SMTP:Wellington.Lacerda@fao.org]
> Sent:	Wednesday, October 04, 2000 3:35 PM
> To:	'tomcat-user@jakarta.apache.org'
> Subject:	RE: Performance
> 
> How can I configure the thread pool for the correct load profile of my
> installation ?
> 
> Are there other known bottlenecks recognised ?
	[Kitching Simon]  

	Note: everything here refers to tomcat3.1. I understand that
	3.2 does things a little different, and 4.0 does them a lot 
	different.

	These issues are addressed in the "real world configuration tips"
	section of the tomcat3.1 user guide. It's pretty brief, but there
	aren't a whole lot of things that need to be done. And what
	the best settings are depends on the web application itself.
	Read the user guide, make a first guess at the settings. This
	will certainly improve the situation over the defaults. After that,
	you really need to do some testing on your app.

	We use Rational load test, it works fairly well but is expensive.
	Perhaps Apache JMeter could be useful...

	I've set up a thread pool of 60 threads, and that looks 
	like it is about right for serving simple jsp pages to 300 
	users hammering it like mad (ie our load test going flat-out).

	The test setup is tomcat3.1 on HPUX11.00, 4-processor 
	PA-RISC2 450 with 1.5GB ram...

	When we've finished our testing, I'll post some conclusions
	to this group about the stability of tomcat3.1. Currently, we're
	getting some problems but haven't pinned down whether it is
	tomcat's fault or not.

	Cheers,

	Simon


> Wellington
> 
> 		-----Original Message-----
> 		From:	Kitching Simon [mailto:Simon.Kitching@orange.ch]
> 		Sent:	Wednesday, October 04, 2000 3:05 PM
> 		To:	'tomcat-user@jakarta.apache.org'
> 		Subject:	RE: Performance
> 
> 		Having had a look at the orion benchmark test, what I
> 		really find interesting is that they have no information 
> 		whatsoever on the *way* that the products were configured.
> 
> 		Tomcat behaves very differently depending on its config
> options.
> 		For example, was the jsp auot-reload on or off? This makes a
> *huge*
> 		performance difference.
> 
> 		And was a thread pool configured? How many threads in the
> pool?
> 
> 		I'm not saying the benchmark is wrong, but that there are
> unanswered
> 		questions.
> 		I see that the JRun people are working with orion to rerun
> the tests,
> 		presumably so
> 		they can tune their webserver properly before the tests are
> run. Until and
> 		unless
> 		a good tomcat person does the same for tomcat, or at least
> orion publish
> 		the config for tomcat, they can't be taken at face value.
> 
> 		> -----Original Message-----
> 		> From:	Lacerda, Wellington (AFIS)
> [SMTP:Wellington.Lacerda@fao.org]
> 		> Sent:	Wednesday, October 04, 2000 2:13 PM
> 		> To:	'tomcat-user@jakarta.apache.org'
> 		> Subject:	Performance
> 		> 
> 		> Have you guys looked at
> 		> http://www.orionserver.com/benchmarks/benchmark.html
> 		> <http://www.orionserver.com/benchmarks/benchmark.html>  
> 		> 
> 		> Why is TOMCAT so slow when compared with the other engines
> ? Which
> 		> performance bottlenecks are recognised and how to avoid
> them ?
> 		> 
> 		> What about Catalina ? is it faster than TOMCAT current
> engine ?
> 		> 
> 		> Wellington Silva
> 		> UN/FAO