Thanks for the reply!

At 12:01 PM 8/3/00 -0700, you wrote:
>Hi Kirk,
>
>I can't help you with your exact problem, but there may be some platform
>issues that you're running in to.
>
> > For these comparisons the test platform is a PIII-550/128MB,
> > Win98, Sun SDK
> > 1.2.2, and Hypersonic as shipped from Orion.  System Monitor shows 100%
> > CPU, as expected.
>
>
>First off, you should probably download the JDK 1.3 -- there's considerable
>performance improvements there.

My concern was relative performance changes, not absolute performance - at 
least, not yet.  We are, alas, some way from production.

The 1.2/1.3 argument is still going around the office.  1.2's fast enough 
for what we're doing now, and gives us a consistent platform.  1.3 seems to 
need to mature a bit on Unix platforms.

>Secondly, Win98 gives major CPU priority to the foreground application,
>unlike NT or unix.  In doing some tests of the orion web server/servlet
>environment on a development box running Win98, I noticed that there would
>be a *huge* difference (up to 10x) in performance if the dos box running
>Orion was in the foreground.  If I brought a different application to the
>foreground, performance would nosedive.

Either Orion or the client is in the foreground, with essentially the same 
results, and no other applications are running.  The tests have been run 
the same way on the same platform all along - they're part of the "does it 
still seem to work" check I try to run on every update.

>Needless to say, I don't recommend running any production servers on Win98.

I'd never consider Win98 for production.  We're a Unix shop, and I've 
admin'd assorted versions for around 20 years.  We haven't made a final 
production platform decision yet.

However, this machine is sitting on my desk so it's real handy for 
development and testing.  It needs to run some win32 applications, and has 
some hardware that NT doesn't like.

One of Orion's strengths is being able to do useful development under this 
pervasive but less than optimal OS.

We ran comparison benchmarks on several OSs, JVMs, and DBMSes, and will 
continue to do so.  This one's the quick and dirty test, and the 3x 
increase in times finally needs to be investigated.  I was hoping the 
problem would go away with newer versions, but no such luck.

>Good luck,

Thanks!  I appreciate it.

>Darren.
>
>--
>Darren Gibbons                                    [EMAIL PROTECTED]
>OpenRoad Communications                               ph: 604.681.0516
>Internet Application Development                     fax: 604.681.0916
>Vancouver, B.C.                                 http://www.openroad.ca


Kirk Yarina
[EMAIL PROTECTED]


Reply via email to