Re: EZ/X Speed: Rebunking Oracle's Debunk: WAS (Getting Some Facts Straight on Performance Claims)

Subject: Re: EZ/X Speed: Rebunking Oracle's Debunk: WAS (Getting Some Facts Straight on Performance Claims)
From: "Michael Sick" <mike@xxxxxxxxxxxxx>
Date: Wed, 08 Dec 1999 10:37:57 -0500
Hi Steve,

> Your latest Jar's benchmark-with-output
> makes the same change I made which was
> to make both try writing out to a file.
> No conspiracy theories. Nothing up my sleeve.

Super. Glad we got it right.

> 
> I'll give this week's Jar a spin when I get back
> from travelling this week at XML99 so I'm
> trying it on the same machine where I ran my
> previous tests for continuity. My initial impression
> is that the 1.2.2 JVM being used makes a big
> difference... I'm not using Sun's default 1.2.2
> in my environment. Maybe I'll download it to match
> your environment.

When optimizing other Java products, Activated has learned that Java
performance can be quite sensitive to both the version and the provider of
the JVM. Things that run quickly in a Sun JVM may lag a little in the IBM
version and vice versa. There can also be significant and non-deterministic
variations due to JIT behavior - we run our tests 5x and take the average to
mitigate these effects.

In the end - it would be great if there was a standard set of xml, xsl,
platform, and JVM combinations that could be used for performance
comparisons.

Regards,

Michael Sick
begin:vcard 
n:Sick;Mike
tel;pager:917 975 6860
tel;cell:917 975 6860
tel;fax:(212) 896-8222
tel;work:(919) 319 3460
x-mozilla-html:TRUE
url:www.activated.com
org:Activated Intelligence
adr:;;1001 Winstead Drive;Cary;NC;27513;US
version:2.1
email;internet:mike@xxxxxxxxxxxxx
title:Vice President
note:Activated Intelligence - Bringing Life To The Net, and The Net To Life
x-mozilla-cpt:;-19184
fn:Mike Sick
end:vcard
Current Thread