As a J2EE sysadmin, I can't think of how running tomcat/apache+mod_jk
on an embedded device without any local storage other than a CF card
and no swap is a good idea at all.  In fact, this is a very bad idea
unless you plan on having a -ton- of RAM and a pretty good processor
to handle all the network i/o + java.  If you do have a lot of ram,
then running the whole jvm inside of an mfs would be cool, but just
remember that you shouldn't write to the CF card very often if you
want it to last.


On Thu, 3 Feb 2005 14:36:48 +0000 (GMT), David Brownlee <abs@netbsd.org> wrote:
> On Thu, 3 Feb 2005, Chris Humphries wrote:
> 
> > they warn use at your own risk. why go to all this trouble to use netbsd
> > when linux is right there and could have had the job done already. i
> > love netbsd as the next guy, but when you got a job to do, you choose
> > the best solution for the task at hand.
> >
> > trying to force a round peg in a square hole may be fun, but when there
> > is a job to do round pegs go in round holes so you can get the project
> > moving and the task done :)
> >
> > many people though have reported java with tomcat on freebsd to run without
> > much problems, so many it may work. just if it was my name on the line, I
> > would choose to go with something that you know works without question.
> >
> > maybe if it wasn't tomcat netbsd vs linux argument would be more sane, but
> > it seems like linux is the obvious answer here.
> >
> > i would suggest playing with netbsd and doing some tests and really stress
> > test the heck out of it if you can get a test environment to do so. while
> > you are at it, do it with linux too.
> >
> > maybe be good for all of us if you did testing and report to us what you
> > find :) if you have a lot of time to test before turning over to the client,
> > i would suggest you test first and make sure you are ok with the choice
> > whatever it may be. if i have learned anything, it is that what i think is
> > going to happen doesn't always happen, so it is better to run tests to
> > experiment and be sure. :)
> >
> > tomcat running on a small embedded device off a flash card seems silly to
> > me. curious to see how well it really works. don't suppose there is any
> > j2me solution that could be made? what are your resource restrictions?
> > is it i386 arch, soekris? Is it just a simple gui interface like that in
> > the firmware on linksys/dlink switch/router/ap devices? Tomcat and java
> > are not that bad in low traffic environments. Hopefully the java code is
> > good enough to help the GC (why having Java programmers learn j2me before
> > j2ee is a great idea, imo [in the same way learning asm before C]).
> >
> > so many questions, yet i have to return to work as well ;)
> 
>         We're running production tomcat 5 under Linux (Redhat EL
>         2 and 3), and NetBSD 2.0 (linux compat), and have been very
>         happy with the result. We had some memory issues with the
>         very early sun-jre15, and switched back to sun-jre14 (not
>         tried again since). Most of our NetBSD boxes are still
>         using the suse73 compat libraries, but we are gradually
>         rolling out the suse91 versions.
> 
>         Our main tomcat internal server is an AMD64 (running in
>         32bit mode), with 2GB of memory and RAID1 raidframe on
>         (unfortunately) IDE disks, so any performance issues tend
>         to be due to postgres sitting in disk wait.
> 
>         We'll move across to a native jre when it becomes readily
>         available, and others have stress tested it :)
> 
> --
>                 David/absolute       -- www.NetBSD.org: No hype required --
> 


-- 
_Matt