HornetQ's fast serialization?

[falcnair]

Have the Infinispan authors looked at HornetQ's persistence API? There is a blog describing it at http://hornetq.blogspot.com/2009/08/persistence-on-hornetq.html

My goal is to make sure my state is not lost if a server (or a cluster) crashes. I think Infinispan could benefit from this.

[manik]

Hi there. Yes, we do work quite closely with the HornetQ guys when developing ideas and concepts. We have looked into HornetQ's persistence and while it is great for spooling stuff on to disk, it doesn't handle the random access patterns of Infinispan very well. Since Infinispan needs to be able to load any arbitrary entry from disk on demand, using HornetQ's spool would mean iterating through the spool from the start to search for the entry.

Using one of the other Infinispan CacheStore implementations would better suit your usecase.

[galder.zamarreno]

We've looked at using the persistence layer created for HornetQ in Infinispan but unfortunately, it does not yet support random access and so we cannot yet use it. For a lengthier discussion check: http://lists.jboss.org/pipermail/infinispan-dev/2009-July/000964.html

Please note that we support different cache store implementations and so, if you wanna make sure your state is not lost, you could configure Infinispan to use a JDBC database based cache store or file based cache store.

We're still working on the documentation but we'll happily help you configure it.