You are viewing a plain text version of this content. The canonical link for it is here.

Posted to users@tomcat.apache.org by Rafael Oliveira de Mattos <ra...@softplan.com.br> on 2015/12/17 21:09:50 UTC

troughput difference

Hello, 

I'm having different results in Tomcat performance using version 6.0.37 and 8.0.23. 

Using tomcat 6 we usually have 
5000 ~ rpm with a response time of 80ms 

Using tomcat 8 the max that we can get is 
2000 ~ rpm with a response time of 200ms 

We are using the same jdk version 1.7.0.80, a nd the same applications on both version, they are installed in the same machine and using APR, but not running at the same time. 
We tried to use the same configuration whenever possible. 

SO: Linux ubuntu 12.04 
JDK: 1.7.0.80 
Anything I should take look at it?

Re: troughput difference

Posted by Rafael Oliveira de Mattos <ra...@softplan.com.br>.

Ok,

Sorry for taking so long to answer, we are using this versions in a production enviroment (one server, to just selected users).

So we solved our problem, we changed the jsp-api that tomcat uses, we are using the jsp-api 2.2 from glassfish (we are still studying the side effects from that change). After we profile, our thread dumps showed that 60% of the time of the request were spended in tag.service methods. At first we thought that the error was in our code, but we didnt change anything in the jsp's. So after searching for the problem, we find this bug report https://bz.apache.org/bugzilla/show_bug.cgi?id=57583. With the glassfish version we are faster then ever, but we afraid of the side effects this can have.

----- Mensagem original -----
De: "Christopher Schultz" <ch...@christopherschultz.net>
Para: "Tomcat Users List" <us...@tomcat.apache.org>
Enviadas: Quinta-feira, 24 de dezembro de 2015 15:28:49
Assunto: Re: troughput difference

Aurélien,

On 12/24/15 4:17 AM, Aurélien Terrestris wrote:
> probably this won't solve your problem but I notice that the random seems
> slow :
> Creation of SecureRandom instance for session ID generation using
> [SHA1PRNG] took [9,870] milliseconds.
> 
> Maybe should you then start by fixing this, it has been discussed many
> times on the mailing list ( securerandom.source=file:/dev/./urandom )

Once Tomcat starts up, it should go fairly quickly. I don't know how
often SecureRandom re-seeds itself (or even if it does), but you only
need entropy during the seeding process. It shouldn't have any effect on
the throughput of a few thousand requests.

> Maybe are there some tests to do to understand if your Tomcat 6 seems fast
> because it uses some caching somewhere, or on the contrary if your Tomcat 8
> is slow because there is a bottleneck.
> I would try one scenario with one small HTML page requested thousands of
> times, and one very big file requested just 2 or 3 times, and compare
> results.
> 
> One thing which could help in understanding, have a look at the status
> servlet which gives statistics (
> http://tomcat.apache.org/tomcat-7.0-doc/apr.html )

Rafael, I'm curious: what is your testing procedure?

-chris

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org