HP-UX vs. NT performace issue

Question

I have a J2EE application that I've recently ported from a WINNT development box
to a 2-CPU HP L-class server. I am running Weblogic 6.1, SP1 on both machines.
On a particularly memory-intensive piece of the application, I've notice a performance
drop of 50% to 75% from the NT box to the HP box.
Each server is running the JDK 131 (w/Hotspot) that comes with the Weblogic install.
I have set all the HP system variables (max_thread_proc, etc.) to Bea recommendations.
To benchmark I used an XML->PDF engine (Apache's FOP) to render a 200 page PDF
document from a static XML-file. No database connection is involved. Below are
the results I see. Initial heap size=max heap size in each case.
HP server (HP-Unix, 2x550 mhz, 2GB ram):
Heap Size=64M, 1 report = 545ms/page (frequent out of memory errors)
Heap Size=64M, 2 reports = didn't try
Heap Size=256M, 1 report = 372ms/page
Heap Size=256M, 2 reports = 1700ms/page
Heap Size=512M, 1 report = 350ms/page
Heap Size=512M, 2 reports = 1675ms/page
Nt dev box (NT4-Worksatation, PIII-933mhz, 512MB ram):
Heap Size=64M, 1 report = 251ms/page
Heap Size=64M, 2 reports = 750ms/page
Heap Size=256M, 1 report = 245ms/page
Heap Size=256M, 2 reports = 500/page
Does this kind of perfomance gap seem normal? I can see a drop in speed due to
the processors. But two reports at once should be faster on the HP box right?
It seems sometimes Weblogic switches over to two processors, but not usually until
I run 3 or more reports at once.
The only difference I can see for sure between the two boxes is that the NT machine
performs at least 10 times as much garbage collection. (Sometimes several times
per page, as opposed to once every 8-10 pages on the NT box.) Garbage collection
occurs a little more frequenly on the HP box when I lower the heap size, but still
not nearly as often as on the NT--at any heap size. Also the HP box runs out of
memory if I lower the heap size.
Is this just some difference in the hotspot implementations? Tomorrow we are going
to try installing the HP-UX native jdk131 from Sun and pointing to that instead
of the Weblogic jdk. But I have a feeling they are the same.
If anyone has any ideas about how I can boost the performance on the HP box, or
what might be causing this gap, I'd be appreciate hearing them very much.
Thanks a lot,
Matt Savino
Quest Diagnostics

Answer 1

Thanks Mike. You were right, I meant to say "once every 8-10 pages on the HP box".
The thing that really bugs me is how much more the HP box degrades when I run
two reports concurrently.
And we are bent on using HP because it is a Corporate Standard. Amen.
So are you saying that JVMs run better on Windows, Solaris, or other forms of
Unix?. I'd be very interested in hearing any general opinions on this.
thx,
Matt Savino
"Mike Reiche" <[email protected]> wrote:
>
Yup, the NT box is faster, looks pretty normal to me. More concurrent
requests
(10+) might be a bit of an equalizer. Why are you bent on using HP?
Your comments regarding who does more garbage collection are not clear
- I think
you wrote 'once every 8-10 pages on the NT box' when you meant HP. I
think you're
saying that the NT box performs more GC than the HP box.
Mike
"Matt Savino" <[email protected]> wrote:
I have a J2EE application that I've recently ported from a WINNT development
box
to a 2-CPU HP L-class server. I am running Weblogic 6.1, SP1 on both
machines.
On a particularly memory-intensive piece of the application, I've notice
a performance
drop of 50% to 75% from the NT box to the HP box.
Each server is running the JDK 131 (w/Hotspot) that comes with the Weblogic
install.
I have set all the HP system variables (max_thread_proc, etc.) to Bea
recommendations.
To benchmark I used an XML->PDF engine (Apache's FOP) to render a 200
page PDF
document from a static XML-file. No database connection is involved.
Below are
the results I see. Initial heap size=max heap size in each case.
HP server (HP-Unix, 2x550 mhz, 2GB ram):
Heap Size=64M, 1 report = 545ms/page (frequent out of memory errors)
Heap Size=64M, 2 reports = didn't try
Heap Size=256M, 1 report = 372ms/page
Heap Size=256M, 2 reports = 1700ms/page
Heap Size=512M, 1 report = 350ms/page
Heap Size=512M, 2 reports = 1675ms/page
Nt dev box (NT4-Worksatation, PIII-933mhz, 512MB ram):
Heap Size=64M, 1 report = 251ms/page
Heap Size=64M, 2 reports = 750ms/page
Heap Size=256M, 1 report = 245ms/page
Heap Size=256M, 2 reports = 500/page
Does this kind of perfomance gap seem normal? I can see a drop in speed
due to
the processors. But two reports at once should be faster on the HP box
right?
It seems sometimes Weblogic switches over to two processors, but not
usually until
I run 3 or more reports at once.
The only difference I can see for sure between the two boxes is that
the NT machine
performs at least 10 times as much garbage collection. (Sometimes several
times
per page, as opposed to once every 8-10 pages on the NT box.) Garbage
collection
occurs a little more frequenly on the HP box when I lower the heap size,
but still
not nearly as often as on the NT--at any heap size. Also the HP boxruns
out of
memory if I lower the heap size.
Is this just some difference in the hotspot implementations? Tomorrow
we are going
to try installing the HP-UX native jdk131 from Sun and pointing to that
instead
of the Weblogic jdk. But I have a feeling they are the same.
If anyone has any ideas about how I can boost the performance on the
HP box, or
what might be causing this gap, I'd be appreciate hearing them verymuch.
Thanks a lot,
Matt Savino
Quest Diagnostics