On Mon, Jun 9, 2008 at 1:51 PM, Andy Spitzer <[EMAIL PROTECTED]> wrote:
> Woof!
>
> On Mon, 09 Jun 2008 12:53:14 -0400, Scott Lawrence <[EMAIL PROTECTED]> wrote:
>
>> Can you quantify that at a system level?  Surely most of them are
>> sharable pages?
>
> Don't have a good gantification, but no, most are not sharable due to the 
> dynamic
> nature of Java object loading.  It's all heap space.  WITHIN the JVM is
> where the sharing comes from, not 'tween JVMs.
>
> Assume 256K per JVM, and it starts to add up fast:
>
> 1. sipXpage
> 2. sipXconfig
> 3. symmitron
> 4. sipXbridge
> 5. sipXivr
>
> As we add more Java into the mix, placing multiple services into one JVM
> becomes more important.  More important, I think, than controlling them
> seperately.
>
> One possiblity of fitting this into the current process based structure
> is to have "proxy processes" that Process Manager would start/stop,
> provision, etc.  Those would do whatever Java control is needed to the
> main JVM to start/stop services.  Those apps would be thin procesesses
> (possibly just shell scripts using sockets)
>
> --Woof!
>


This would point to a solution that uses an application container such
as J2EE  to build such services but that would involve significant
re-engineering ( not achievable in the near term ). Such a container
would provide for application isolation using classloaders and
management facilities and so forth.

Ranga




-- 
M. Ranganathan
_______________________________________________
sipx-dev mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-dev
Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-dev

Reply via email to