cache stores / loaders in Distributed mode

[camechis]

I am investigating using infinispan in distributed mode.  I am trying to understand how the loaders/stores work in this mode. If you have 5 nodes do they each write back to the store?  Is there one node that acts as the loader/store?  Is it smart enough to know if the data is replicated on 2 of the nodes to only store one of them?  I am Mainly just trying to figure out how to step cache stores/loaders in this mode ( store probably being a jdc store/loader).

Thanks

[camechis]

Any info on or links on how to properly setup a cache store and loader operating in Distributed mode?

[sannegrinovero]

Other than this http://community.jboss.org/wiki/CacheLoaders ?

[camechis]

yeah, just wasn't clear on how it works in distributed mode.  Does each server node have a cache loader / store that writes back but then what happens if you have data that is replicated on 2 of the nodes for redudancy, does it get saved twice or is it smart enough?

[galder.zamarreno]

Well, it all comes down to whether the cache store is shared or not. If it's shared, only the node where the call originates will store it. If the cache store is not shared, each node has its own cache store, so the originating node and the nodes that store copies will update their cache stores.

[camechis]

I see, just to make sure I am clear, A shared store would be something like a jdbc store correct?

[brenuart]

Yes, although conceptually a fileStore could be shared as well if located on a NAS or else... but I wouldn't advice to do so.

[galder.zamarreno]

Exactly, a typical shared cache store would be a JDBC database, or an Amazon S3 store...

FileCacheStore is better suited for unshared stores IMO.

[mircea.markus]

+1 to both points.