Howdy. My application requires clustered invalidation, so JBoss TreeCache seems like the obvious choice. The only problem is that there are only two available options for caching strategies on Hibernate objects: read-only and transactional.

"Hibernate In Action" advises one to use transactional caching judiciously and only where needed. My problem is that most of my objects will benefit from clustering, and in an OSCache or EHCache environment, they do. But I get a performance penalty from using "transactional" caching. The only other option is "read only" but only some of my objects are truly read-only. Many of them change infrequently, but they do change.

What is the best approach in this scenario? Lots of objects that can benefit from caching: 20% read-only, 80% write-infrequently. The best model for this would be nonstrict-read-write, but that's not available with JBossCache.

Not sure what happened to my Subject line (brain deadness on my part, no doubt). Anyway, it should read "JBossCache - Why no read-write mode?"

Won't there be a performance penaly if i try to use TreeCache or any other cache for that matter, as second-level cache in a CLUSTERED environment?


For example, i've three nodes in the JBoss cluster and assume that the possibility of my primary node coming down is rare. say once in 4-6 months. In that case, if i choose to have TreeCache as second level cache for hibernate and use cache for 20% of my objects in a large scale application, TreeCache will try to replicate the data on the other two stand-by nodes. Won't i pay the performance penalty for this?

Does any one have data on what's the difference in performance WITH and WITHOUT clustering while using TreeCache as second-level cache in Hibernate?

Thanks!