 |
JRockit 5.0 on desktop - show me the money
Posted by kirillcool on December 13, 2005 at 08:24 AM | Comments (9)
In the world of seemingly arbitrary records set by JRockit 5.0 JVM from BEA (see this comment on the JavaLobby thread), i have decided to take my test Swing application for a ride.
This application is a test-bed for Substance LAF, so it utilizes a lot of Swing-related stuff, creating all possible core components, a lot of listeners, various layouts, colors etc. After downloading the latest release of JRockit 5.0 and installing it, here are the results (first of all i should mention that i didn't get any compilation or runtime errors, so i guess that "write once run everywhere" is still partly relevant).
There are two phases that i had timed - the UIManager.setLookAndFeel function (which eventually calls the constructor of some LAF, but also does a lot of other stuff) and the creation of the main frame itself (constructor, pack, setting the size, location and visibility). The results are:
- For JRockit R26.0.0 version 5.0_04 - 1890 ms for initializing the LAF, 3690 ms for showing the main frame and 22MB taken on the heap.
- For Sun VM version 5.0_06 - 1200 ms for initializing the LAF, 1430 ms for showing the main frame and 7MB taken on the heap.
- For Sun VM version 6.0 b63 - 840 ms for initializing the LAF, 1500 ms for showing the main frame and 8MB taken on the heap.
The results are for single-CPU (2.4GHz) running WinXP SP2.
As usual, the results of such a 'benchmark' should be taken with a shovel of salt, since you may have different LAF and different Swing application altogether. However, the main conclusion is that JRockit may be 'smoking' hot for running the WebLogic, but it's certainly not a prime option for running desktop applications.
Bookmark blog post:  del.icio.us  Digg  DZone  Furl  Reddit
Comments
Comments are listed in date ascending order (oldest first) | Post Comment
-
For an application that will be used throughout the day, the start up time isn't usually the major performance concern. How long does opening a second main frame take?
Posted by: tackline on December 13, 2005 at 10:19 AM
-
tackline,
That's a difference between what you expect from a desktop application and from an application server. When you click on an icon (let's say batch file that runs the jar) on desktop, you expect something to happen. If it takes 5.5 seconds, i'll start thinking "may-be i missed the double-click, let's click it again (and again)" and you'll end up even more frustrated when 2-3 instances of your application will eventually open.
You say what's important to you - that's why the word "shovel" is there. I'm not really going to compare these VMs on every possible UI scenario - that's what VM people are paid to do :) In addition, i'll then have to compare all of them (including Apple VM and IBM VM) on all supported platforms - something that you are welcome to do.
Posted by: kirillcool on December 13, 2005 at 09:31 PM
-
JRockit needs some work even on the server side. When we connected up Weblogic with MQSeries, JRockit core-dumped, while the Sun jvm worked fine. Took us a couple of hours to find the root of the problem. I don't know whether there was a specific conflict with the MQSeries JNI code or if this extends to JNI in general. So I would say: make it work first, then make it fast :-)
Posted by: tobega on December 14, 2005 at 01:43 AM
-
Where you running Sun VM with -client or -server? Is there any difference?
Posted by: sla on December 14, 2005 at 04:12 AM
-
About the MQSeries problem: this is actually documented as a known issue (with a workaround) in the JRockit release notes (look for "MQSeries").
Posted by: sla on December 14, 2005 at 04:27 AM
-
sla,
On -server setting under Sun VM 5.0_06 it takes 1700 ms to initialize the LAF and additional 2500 ms to show the main frame.
Posted by: kirillcool on December 14, 2005 at 04:29 AM
-
sla,
Thanks, that solved it. Funny that BEAs paid-for support didn't know that :-)
Posted by: tobega on December 14, 2005 at 06:18 AM
-
JRockit or IBM JVM. Both hurt Java and the platform independence.
Are you aware that IBM JVM 1.4.2 Windows download from IBM website runs on IBM hardware only?
As long as Java is governed by Sun, the "big" players in the Java space will try to hijack Java (remember MS?).
Posted by: gullet on December 15, 2005 at 12:03 PM
-
> JRockit or IBM JVM. Both hurt Java and the platform independence.
No, both help java and platform independance a LOT. IBM's Java runs on many platforms SUN simply does not support (AIX/PPC, Linux/PPC, os360/90, ...) and JRockit supports jit-caching and has other benefits.
> Are you aware that IBM JVM 1.4.2 Windows download from IBM website runs
> on IBM hardware only?
Yeah for sure. Its their god givien right to decide to whom they give the JVM. I would not bother even if you would have to charge for since if you need it you would be happy that it exists and pay.
They invest tons of money into their runtime so why should they be not allowed to get some back with hw.
Furthermore you do not have any reason to be concerned, SUN jvm does exist on windows. Furthermore IBM-JVM is server-side and how really plans to use Windows for serving high-load hava ;)
Posted by: linuxhippy on December 15, 2005 at 11:10 PM
|
java.net RSS Feeds
