If you run top, and make a servlet to call a System Runtime.  Then, use
Jmeter to hammer it.  You will see the swap disk is create the same size as
the -Xms for each thread.  I have done the test with different sizes of
heap.  It is solid true.

Billy Ng

----- Original Message -----
From: "Yoav Shapira" <[EMAIL PROTECTED]>
To: "Tomcat Users List" <[EMAIL PROTECTED]>; "Billy Ng"
<[EMAIL PROTECTED]>
Sent: Tuesday, June 17, 2003 4:40 AM
Subject: Re: heap size config


> Howdy,
>
> >In Unix, java will make
> > 512MB swap every time we make a System Runtime call.
>
> I don't think so: can you prove the above?
>
> >Will this
> > hurt the Tomcat preformnace?
>
> Lowering -Xms will only slightly degrade performance during the allocation
> phase.  The effects are negligible during that phase and none afterwards.
>
> >I read the article from Sun saying the -Xms and
> > -Xmx should be the same.  Is it always true?
>
> No.  Almost nothing in the realm of performance tuning is always true.
Setting
> -Xms equal to -Xmx is useful only if you're going to have a constant level
> memory usage more or less throughout the life of your JVM.  If you start
low
> and stay low until peak demand comes in, keeping -Xms low is more
beneficial.
> GC and other operations are always faster on a smaller heap.
>
> Of course, nothing is more beneficial then having stress tests to run
against
> different combinations of settings.  That way you can see the actual
effect on
> your system intead of relying on me or someone else's article.
>
> Yoav Shapira
>
>
>
> =====
> Yoav Shapira
> [EMAIL PROTECTED]
>
> __________________________________
> Do you Yahoo!?
> SBC Yahoo! DSL - Now only $29.95 per month!
> http://sbc.yahoo.com
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to