> c- we write a trivial CMP engine, in fact we don't write we plug
> in your distributed DSS thingy from the clustering UNDER the CMP
> store API and voila we have an entity semantic to find the stuff
> with your hashtable under the covers
>
> If this performs well, we will have proven one of the biggest
> points of EJB in JBoss.  All it takes is putting that DSS in a
> CMP API wrapper.

I've done it yesterday:
        - new PersistentStore backed with DSS (I've also commited a pure in-memory
PersistentStore... just in case...)
        - new InstanceCache with distributed semantic i.e. if a bean modified on a
node, all other nodes caches are cleared (of this particular bean)

Now I will write the simple EJB with a dedicated stack (that uses these new
layers) and possibly a helper JMX service (or simply some helper classes)
that will simplify/wrap the usage of the bean. The helper service/class will
also provide a mean to subscribe for cache update. Consequently, if Julian
wants to keep a cache of HTTPSession, it can do it very easily witout
loosing coherency. And this would really speed up things because as long as
the cache is ok, there is no EJB=>HTTPSession (i.e. byte[]=>class)
serialisation work to perform at all.

Your dream is comming reality ;)

Cheers,



                                Sacha


_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development

Reply via email to