tek1 wrote:
>
> At 17:16 03/05/30 +0000, you wrote:
> >This one time, at band camp, tek1 said:
> >
> >t>Just read an interesting article about performance improvement using JBoss,
> >t>which I am using with Castor.
> >t>
> >t>The article was focused on EJB, but the author talked about a "Cache
> >t>Invalidation Bus" to notify the cache of other JBoss instances that a
> >t>particular entity was modified (in another JBoss instance) and that the
> >t>entity should be removed from the cache of the other JBoss instances.
> >t>
> >t>http://www.onjava.com/pub/a/onjava/2003/05/28/jboss_optimization.html?page=2
> >t>
> >t>
> >t>Perhaps a similar concept could be adopted for clustering in Castor?
> >t>
> >t>Is anyone currently using Castor in a clustered environment?
> >
> >JBoss accomplishes via JavaGroups, a project by Bela Ban. Actually
> >I've been pondering this with my cache refactoring ideas and I've
> >been meaning to speak with Bela about it.
> >
> >My idea is to provide some default caching mechanism but make this
> >layer pluggable so that any caching mechanism could easily be plugged
> >in, including a distributed cache. This requires the use of a well
> >known API (e.g. possibly JSR 107 via http://jcache.sourceforge.net/,
> >http://sourceforge.net/projects/ojc or something similar) so that this
> >layer can be pluggable. But these two projects are in their infancy so
> >I'll need to speak with Bela about JavaGroups and what API he implemented
> >for it.
> >
> >Bruce
>
> JavaGroups looks interesting.
JavaGroups is nice, I used it for writing a simple fail-over agent for
an internal project and it works quite nicely.
>
> So would each JDO (and its associated cache) instance have a uniquely
> identifiable name for its "Group" in JavaGroups, perhaps IP address +
> counter (or other id/number that could uniquely identify several JDO
> instances on the same machine)?
I'm not even sure the individual names matter in this case. You'd just
broadcast the message to invalidate a given object. Any Castor JDO
instance listening in on a particular group-channel would then honor the
request and send an acknowledgement.
>
> Then, when one JDO instance ("Group") has a value in its cache change,
> notify the other JDO instances via JavaGroups?
>
> Also, I asked in a previous post, but has any thought been given to
> releasing 1.0 of Castor and then working on the clustering/distributed
> cache for version 2.0? Even though Castor is perhaps one of the oldest
> object-relational mapping frameworks around, some may be hesitant to use it
> given that it has never reached "1.0" status. It's probably more
> psychological than anything else... ;)
Yes. Once we hit Castor 0.9.6 we'll be on the fast track to 1.0. My goal
is for 0.9.7->1.0 to be only bug fixes, additional test cases, and
documentation. No feature develepmont. 0.9.6 will be the last release
with new features, which is why we're moving slowly to that point
because there are still features we're trying to finish up.
Actually we're ready to do a new release, so hopefully I'll find the
time this weekend to do it.
--Keith
-----------------------------------------------------------
If you wish to unsubscribe from this mailing, send mail to
[EMAIL PROTECTED] with a subject of:
unsubscribe castor-dev
