Multi site and replication

Why do you need to replicate? There are simpler alternatives.

1. You could use a SAN (amazon ebs) to persist the data, with load balanced horizontally scaled SPARQL nodes. 4store, Virtuoso, Fuseki and pretty much any triplestore should support this model.

2. Depending on how often data changes, you could alternatively use a publisher, subscriber model following a star topology for updating data.

3. Alternatively, look at using SAAS. I am currently using a beta of Dydra and am finding it quite useful (don't have to worry about the underlying infrastructure). Remember, it is only a beta and might not be production ready, plus you need to sign up for an account.

A rather light replication protocol can be achieved using PubSubhubbub, a publication/subscription protocol : store B subscribe to the update feed of store B and gets RDF data which get merged in the local store.

Some references :
- SPARQLPuSH in this solution, the PuSH feed is genertaed through SPARL Queries
- SDShare
- older idea : RDFSync