suggestions for a replicated datastore

I am looking for a simple key-value datastore which will automatically replicate itself on different machines. Unfortunately, a Dist开发者_运维知识库ributed Hash Table will not work for me since I need the whole datastore to be available on all machines. I have looked at mnesia from the erlang world but talking to it from different languages is a pain.

Any suggestions on what I should go for?

Thanks!

---

DNS. It's a distributed cached mapping table replicated across all masters (and partially to clients).

I think we can safely say it scales.