tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From André Warnier ...@ice-sa.com>
Subject Re: deployed same TOMCAT 6.0.20 on Windows / Linux
Date Wed, 09 Dec 2009 16:08:26 GMT
Karthik Nanjangude wrote:
...
> 
> When something is in Production it really gets pickles to upgrade the Jdk /OS , Realy
worried ....  :(
> 
Karthik,
generally speaking, the gurus on this list - of which I am not - tend to 
be sceptical about benchmarks.  That is because, as Mark indicated in 
another thread a couple of days ago, you can make a benchmark that will 
tell you anything you want.
It is incredibly difficult to create a real benchmark that will really 
tell you something worthwhile.
The best benchmark you can have, is to use your own application, in 
conditions as close as possible to production, and see how it behaves.

Your previous posts indicated that you were doing some kind of test, 
presumably to obtain some valid information that you can then use to 
deploy an application on one server rather than another.
To get such valid information, you must take a number of precautions, 
otherwise your results will be nonsense, and will ultimately lead you to 
take the wrong decision.

What the various people answering you so far have tried to do, is to 
tell you that, in their more or less expert opinion, the conditions in 
which you are doing these tests now, according to the data you provide, 
do not so far look as if you would get valid results out of them.

Tomcat runs inside of a Java JVM, which in the principle should isolate 
the Java program (in this case Tomcat) from the underlying OS.
There is no reason /in principle/ why a Tomcat application would run 
slower under a Windows OS than under a Linux OS.
But there are so many external dependencies, like : the hardware, the 
memory available to Java, the Java version itself, your command-line 
switches to start Java, the network, what else is running on the 
machine, and so on, that this kind of comparison is bound to disappoint 
you.  And testing with a HelloWorld JSP page, is in no way comparable to 
what will happen to your real application and real server under load.

There exist tools that will issue automatically a number of requests 
simultaneously and over a period of time, to a webserver, and that will 
produce nice results, sorted in a table, easy to read.
You may be able to use them to issue real requests, to your real 
application.  That would be a much better test.
But it will still not compensate from testing on two different servers 
which, on the face of it, look like they are quite different, even 
abstracting the OS.

And because you are using quite old versions of software, it is unlikely 
that anyone of the few people on this list that would be willing to 
help, would actually be able to do so.  That is because thay could not 
check any result that you have, with a comparable system that they have.



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


Mime
View raw message