Originally Posted by Justin Albstmeijer
I wonder if there is a global design overview somewhere....
No. But the setup is really simple. One master database and a unlimited number of slave databases which replicates only the data needed for a psecific server.
The reasons for this setup are:
1) The access to the local DB is much faster and can also use sockets.
2) You would reach a max. connection limit very fast if you had just one central db.
3) Security: the slave databases contain only the data for the local node.
we started with a central Db setup in the first betas of ISPConfig 3 and the approach was not scalable enough. The current setup has been implemented by several providers already in production systems and it works very well.