geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Matt Hogstrom <m...@hogstrom.org>
Subject Initial Performance Measurements of Geronimo Using DayTrader - Bottom Line ... we're gonna mix it up :)
Date Fri, 07 Oct 2005 19:58:31 GMT
I have performance tested M5 (or thereabouts) using DayTrader (a performance 
sample in Geronimo currently located in the sandbox).

The results of the testing I think are quite compelling as Geronimo just passed 
CTS certification at M5.

I used a metric called a "Target" in the testing.  This metric is comprised of 
Open Source and Commercial AppServers.  As we make changes to improve 
performance when we meet the "Target" we will be competitively positioned.

The WebContainer primitives are a bit misleading as they are really contrasting 
Jetty (Geronimo's default) Tomcat and commercial offerings.  We need to fix 
Geronimo to improve performance by allowing HTTP logging to be disabled.

As we move up the stack and include higher level functions (like DB read and 
writes) we find that Geronimo is well positioned against the target.  Prepared 
Statement caching will move us to close to the target I expect.

The EJB primitives may seem poor at the moment.  Although, my gut tells me that 
we are copying parms inappropriately and we'll make  significant performance 
improvements as we address this issue.

Overall Geronimo is within 70% of our competitive target for Trade Direct 
Performance which is awesome.  This metric includes Servlets, JSPs, EJBs and 
messaging to name a few of the J2EE artifacts.  At this point in the cycle we 
should be able to close the gap and make our competitive target by 1.0.

Attached is a set of slides that explains the testing, benchmark and results. 
Please take a few minutes to review and hats off to the team so far.  We are 
going to make Geronimo a force to be reckoned with by 1.0 which is coming soon.

- Matt

Mime
View raw message