Like any well planned out software project the ideal solution is derived
from the requirements. There are very few broad statements such as "Maybe
for large volume servers, but for smaller servers (<10 requests per second)
it is NOT WORTHWHILE to do something else than Tomcat standalone" that hold
true over the course of time.

Is the only requirement for a web project the # of users and traffic? I have
yet to come across such a project, and I've been at this for quite a long
time. What are the security requirements? Do you have any legacy cgi code
that tomcat can't handle well? Will you need to add any? Are you sure?

I'm not sure why people are so afraid to configure up apache. If you can get
tomcat configured you most certainly can get apache running. It's much
simpler (not that tomcat is hard). 9/10 times apache can be configured and
running on a box in about 5 minutes if you know what you're doing. I'm
currently installing tomcat/apache for a new intranet. The requirements
didn't start out necessitating apache (based on load and feature requests),
but I thought that I might as well just in case. There's a lot of static
images, why bother tomcat with those tasks. In the back of my mind I knew
that something would creep in. Sure enough, just today we decided to try to
authenticate users against the NT domain. I'm sure someone out there is
going to tell me how to do that from tomcat, but it will probably require a
bunch of new code and be a major hassle. Guess what, Apache has a module for
it (and I'm running on solaris). 

IMHO the software industry is full of stories about packages that died
trying to do too much. OO programming is all about modularizing code.
Doesn't that transfer to server software as well? Tomcat is a servlet
engine. Apache is a webserver. Sure tomcat can serve .gifs, and I could
write a novel in Outlook, but I'd rather use a word processor. If we push
the tomcat engine towards being a fully featured webserver isn't it
eventually going to bloat? Won't the task we asked it to do (handle
servlet/jsp requests) wind up suffering?

Anyway, that's just my opinion. 

Rick


-----Original Message-----
From: Peter Mutsaers [mailto:[EMAIL PROTECTED]]
Sent: Thursday, October 18, 2001 7:39 PM
To: [EMAIL PROTECTED]
Subject: Re: USE TOMCAT INSTEAD OF IIS


>> "Rui" == Rui Miguel Seabra <[EMAIL PROTECTED]> writes:

    Rui> No.

    Rui> Use apache web server instead of iis or tomcat webserver.

    Rui> Even the configuration is MUCH MUCH easier (and safer even in
    Rui> default config).

Maybe for large volume servers, but for smaller servers (<10 requests
per second) it is NOT WORTHWHILE to do something else than Tomcat
standalone.

It is way easier to configure tomcat standalone than configure two
programs (tomcat and apache) plus the connector between tomcat and
apache. 

Also, if most of your pages are mostly dynamic, relaying everything
through apache will only slow down compared to direct access to the
servlet engine (tomcat).

Only for high volume static pages I would recommend (maybe) adding
apache.

Maybe the optimal would be to have two servers, one for dynamic stuff
and one for static stuff (the images etc). The dynamic pages refer to
the other server (apache) for larger static data. Thus you get the
best of two worlds: direct access to the servlets without
deviation/routing through apache, and faster(?) access to static data
through apache. B.t.w. SUN claims a java webserver/servlet engine
should be able to reach the same performance for static data as
dedicated webservers, but I haven't done any benchmarks.

I can only say that after using Apache+Tomcat for our (internal) app
for 10000 users at about 20 requests per second, we ditched Apache and
went to Tomcat (version 4) standalone. This simplyfied configuration
and improved performance.

-- 
Peter Mutsaers  |  Dübendorf    | UNIX - Live free or die
[EMAIL PROTECTED]      |  Switzerland  | Sent via FreeBSD 4.3-stable

Reply via email to