|
Replies:
6
-
Last Post:
May 2, 2005 1:05 PM
by: boaz_bagbag
|
|
|
|
|
|
|
Test against third party applications
Posted:
Apr 28, 2005 10:45 AM
|
|
|
Is the Sun team testing the Mustang against third party applications. I know Microsoft has used this technique with their products. For example, it would be nice to see if their can be tweaks with Resource intensive apps like Eclipse, Tomcat, JBoss. I know you can't just make changes for each and every application, but the major ones it would be nice two see targeted.
And, if you want to know, Eclipse with Mustang b33 seems to Run faster, the memory is still above the 100MB mark. But, that is normal.
|
|
|
|
|
|
|
Re: Test against third party applications
Posted:
Apr 28, 2005 6:22 PM
in response to: blbrown
|
|
|
Hi,
> Is the Sun team testing the Mustang against third
> party applications. I know Microsoft has used this
> technique with their products. For example, it would
> be nice to see if their can be tweaks with Resource
> intensive apps like Eclipse, Tomcat, JBoss. I know
> you can't just make changes for each and every
> application, but the major ones it would be nice two
> see targeted.
We test a wide range of internal and external applications on a regular
basis. We test appservers, webservers, portals and other large Java
applications that are used in enterprise environments. The diverse set
of applications helps us assess the quality of our releases before they
are delivered.
Here are some examples of Sun or third party applications
we currently use to test the JDK from various different
aspects (reliability, stability, compatibility)
tomcat, netbeans IDE, S1 appserver, etc.)
In addition we track the performance characteristics
of every build that is "promoted" using a specialized
perfomance measurement lab and associated infrastructure.
Keep in mind that prior to the beta and FCS releases
a lot more longer period of testing is done
(affectionately known internally as the "baking period").
Stay tuned, I talked with an engineer in the Quality
group who said that she might do a blog about
the JDK internal testing process !
Thanks,
Bino.
|
|
|
|
|
|
|
|
|
Re: Test against third party applications
Posted:
Apr 28, 2005 6:26 PM
in response to: bino_george
|
|
|
I posted this question on a javalobby forum and this is another response I got. Thanks for the reply. And my next question, do you think part of the test suite will be available for download for mustang?
"We do a lot of that. There's a special test suite, for example, devoted to compatibility with MSVM, which consists of hundreds of popular applets from all over the web. And believe me, those compatibility bugs are very high priority.
And there are many other test suites, both client- and server-specific (for example, one called BigApps, which includes a bunch of heavyweight server apps).
There are suites which run continuously, just to see for how long a bunch of apps can be run (of course, with simulated load). This suite runs for weeks.
There's a performance suite (called Alacrity), very sophisticated, which is used to monitor many aspects of performance (footprint, startup, runtime performance).
And, of course, we have huge regression test suits (not sure why the j2se's suite isn't included with the j2se mustang workspace, it's a part of it, ~10000 tests that I just counted).
Most of them are automated, some aren't.
Would be worthwhile to push some of our sqe folks to start blogging and explain the testing processes we have inhouse."
Thanks,
*Name Removed*
Java2D Team
|
|
|
|
|
|
|
|
|
Re: Test against third party applications
Posted:
Apr 29, 2005 10:42 PM
in response to: blbrown
|
|
|
Hello, I'm the "Name Deleted" from the javalobby thread. =)
You asked there:
> And, I am sorry, but I have to ask. I notice a plain
> Jane java application carries a pretty light load with
> maybe a 5MB-6MB memory footprint and normally runs
> pretty fast. But large apps like Eclipse, JBoss and
> Tomcat eat up an enormous amount of resources. I know
> mustang will include a server vm which may solve a lot
> of resource issues. Where do you think all the memory
> Issues are coming from. And, I guess it may not be an
> issue. Eclipse for example, I have never seen under
> 110MB, Tomcat can eat up 30MB on the simplest servlets.
> Speed normally hasn't been a big issue, but the memory
> footprint has always concerned. And, who knows it is
> probably programmer error.
It was kind of off-topic on the javalobby thread, and that
thread was getting pretty old, so I'll answer here.
The question of 'where that memory problems are coming from' is one of the main interests of special team we have (which actually consists of members from many teams across j2se. They also look at startup time problem, and most of startup improvements in jdk 5.0 were the results of this team's activity.
One of the testing frameworks we have tracks the footprint of many client and server applications, and regressions in footprint are tracked at high priority.
It'd be impossible to answer a generic question about the footprint, though.
There are many aspects contributing to it, including the vm, libraries problems, and not optimally written applications.
Surprisingly, even measuring footprint is a very non-trivial task. Which of the footprint characteristics are more important, actual footprint or perceived one (the one showed in most widely used tools to look at processes' footprint, like top or task manager)?
I guess what I'm saying here is that I can't give you the precise answer, but we're really working hard on fixing the footprint issues.
Again, I'm hoping to convince some of the folks on the team to blog about our efforts in this area.
Thanks,
Dmitri
Java2D Team
|
|
|
|
|
|
|
|
|
Re: Test against third party applications
Posted:
Apr 29, 2005 11:03 PM
in response to: trembovetski
|
|
|
We live in a small world don't we.
|
|
|
|
|
|
|
|
|
Re: Test against third party applications
Posted:
Apr 29, 2005 11:10 PM
in response to: blbrown
|
|
|
One more note. If you use the non-scientific tools like the Windows Task Manager and "top", a simple java application can run with at 6MB with optimal speed which is pretty low even for a windows console app. I don't know how some applications can eat up to 20 times that much.
Right now with the "Task Manager" 'JEdit' is pretty low with
25MB.
I just starting using a tool called UMLet which is a swing application that runs at 25MB(where 25 is a good number)
And on the other side, applications like FireFox are pretty hefty, for example now it is running at 40MB
This is on WinXP 2ghz - 512MB.
I think it is a matter of "how" to write good Swing apps.
|
|
|
|
|
|
|
|
|
Re: Test against third party applications
Posted:
May 2, 2005 1:05 PM
in response to: blbrown
|
|
|
.
|
|
|
|
|
java.net RSS
