C-JDBC and clusters

Anyone use?

I am into month 4 of a detailed study expidention into making business
(SME) clusters at an effective price with no juggling and piss farting
about.

So far I have come to these conlcusions:

If I can keep all my applications as database centric as possible (Even
POP AND SMTP) it is much easier to scale out and backup/migrate etc.

Now clustered file services are easily done with a shared external SCSI
housing and a NFS failover or GFS setup.  However database clustering is
harder.

MySQL's database app only has memory based clustering so no go there.
Mysql also does offer replication.....

I have a situation where I want to be able to use many replicated
databases - easy with MySQL if I know that one transaction will never be
run on more than once database at a time.

C-JDBC looks intriuing for those apps that are written in Java.  I guess
I need a way to cluster a database easily for speed and fault tolderance
- that can be scaled out with all nodes of the database being equal.
That is - I would like to have the cluster as flat as possible without
any "super nodes".  

A "Super node" cluster is fine for medium to larger clusters and makes
sense.  But I would like a model that uses the same nodes as the lowest
common denominator and then scales out with the same hardware.  With a
central node being able to be seamlessly supplanted by a super node as
the cluster grows and the previous central node becomes a normal node.

Are my req's getting too complex?  Is it possible to have cheap small
clusters that just work?

Cheers,
Richard.

_______________________________________________
blug-general list
[email protected]
http://list.beijinglug.org/cgi-bin/mailman/listinfo/blug-general