|
| Re: EMF memory consumption details [message #429618 is a reply to message #429614] |
Mon, 27 April 2009 16:43   |
 Ed Merks
Messages: 33140
Registered: July 2009 |
Senior Member |
|
|
This is a multi-part message in MIME format.
--------------060109090904000809090609
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit
Vincenzo,
Comments below.
Vincenzo Caselli wrote:
> Hi all,
> in Eike Stepper's recent talk at JAX 2009 in Mainz, Germany, he
> mentioned that EMF shows some scalability problems (not releasing
> objects in memory).
Resources are demand loaded and there's nothing that will allow any of
the loaded content to be collected unless you actively take steps to
unload the resource...
> We would like to know more details about this, since we definitely
> believe in (and counting on) EMF Modeling infrastructure for most
> projects.
Really, you need to measure.
>
> Which guidelines should we follow in order to avoid this?
It all depends on what you're trying to do. Generally editors don't
work with instances that are so massive that they'd use up all available
memory, since users can't generally cope with so much information.
> How the combination CDO+EMF behaves with respect of this issue?
Eike's blogs have nice details:
<http://thegordian.blogspot.com/>
http://thegordian.blogspot.com/
Be sure to check out the ones from November.
>
> Thank you very much.
>
> Vincenzo
--------------060109090904000809090609
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Vincenzo,<br>
<br>
Comments below.<br>
<br>
<br>
Vincenzo Caselli wrote:
<blockquote
cite="mid:df6c170e102abce2930dafb175415723$1@www.eclipse.org"
type="cite">Hi all,
<br>
in Eike Stepper's recent talk at JAX 2009 in Mainz, Germany, he
mentioned that EMF shows some scalability problems (not releasing
objects in memory).
<br>
</blockquote>
Resources are demand loaded and there's nothing that will allow any of
the loaded content to be collected unless you actively take steps to
unload the resource...<br>
<blockquote
cite="mid:df6c170e102abce2930dafb175415723$1@www.eclipse.org"
type="cite">We would like to know more details about this, since we
definitely believe in (and counting on) EMF Modeling infrastructure for
most projects.</blockquote>
Really, you need to measure. <br>
<blockquote
cite="mid:df6c170e102abce2930dafb175415723$1@www.eclipse.org"
type="cite"><br>
Which guidelines should we follow in order to avoid this?
<br>
</blockquote>
It all depends on what you're trying to do.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
| Re: EMF memory consumption details [message #429621 is a reply to message #429614] |
Mon, 27 April 2009 19:01 |
|
Vincenzo,
As Ed mentioned, much of this background is explained in
http://thegordian.blogspot.com/2008/11/how-scalable-are-my-m odels.html
To not be mistaken: I don't think that EMF has more issues than almost
any hand-written model here!
One of the underlying problems is containment which always references
children with strong references.
In addition the Resource.unload() method that Ed mentioned as well, does
not necessarily do what you want.
Usually it does not reduce memory footprint. The opposite is the case,
since the proxy URIs take up additional space!
Finally, CDO (in combination with EMF) solves all of these scalability
issues, as explained in the blog ;-)
Cheers
/Eike
----
http://thegordian.blogspot.com
Vincenzo Caselli schrieb:
> Hi all,
> in Eike Stepper's recent talk at JAX 2009 in Mainz, Germany, he
> mentioned that EMF shows some scalability problems (not releasing
> objects in memory).
> We would like to know more details about this, since we definitely
> believe in (and counting on) EMF Modeling infrastructure for most
> projects.
> Which guidelines should we follow in order to avoid this?
> How the combination CDO+EMF behaves with respect of this issue?
>
> Thank you very much.
>
> Vincenzo
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
Powered by
FUDForum. Page generated in 0.02992 seconds
] 
Search
Help
Register
Login
Home
