http://c-jdbc.objectweb.org

I've been starting to check this out and it seems to be very interesting.

Does this mean that Hibernate can now scale as much as it wants ?

--b

Please don't make assumptions about Hibernate. It's a clustered JDBC driver, one layer below Hibernate. Yes, it is nice, but good use cases are exotic.

ok then, let me rephrase that ...

do you think it would enhance hibernate ?

--b

I think its a good and useful idea, but it seems to me there have to be serious transaction-related issues to consider (either isolation is compromised, or you need some kind of 2-phase commit, perhaps even consistency between nodes is not guaranteed).

Remember, in distributed systems you have a choice between guaranteed delivery or 2PC, neither of which seem perfect here.

More generally, does it enhance Hibernate? Well, yes, as much as it enhances scalability of any app that uses JDBC...

Yes, the two phase commit is something that i have allways wondered about
with clustering.

Does time to save and update allways increase exponentially as new
nodes are added to the cluster of whatever it handling persistance ( ie
app servers/c-jdbc etc ).

I guess what I'm asking is this....

Is it better to just pay up for a kick ass database that transparently
handles the scalability or is it better to pay for/invest your time in
java clustering software ?

Seeing as two of the main java persistance dudes ( ie Gavin & Christian )
are in this thread it would be interesting to hear their opinions.

--b

I think it is better to buy solution from RDBMS vendor and to buy adapters to support "old" clients.
If you have a farm of different databases and you need stuff like replication and clustereing then probably you have many clients, different conectivity and programming languages in system. If you have money to maintain this farm then probably you have money for cleanup too.
I do not want to recommend any vendor or free database, it is not trivial decision and I think there is a big risk to fail in this kind of migration.