I am looking into the possibilities of how to scale out the sql server 2008R2. simply scaling up wont be enough.
"p2p replication" and "distributed partitioned views" look interesting, but both require (multiple?!) enterprise licences which are very expensive, and still both arent perfect solutions.
One very simple Idea was:
Read from Random Server - Write to all Servers
Nicely wrapped in a transaction so that it could be rolled back if it doesnt succed on all servers .I am having fear of some serious locking issues if something goes wrong on one server or it is waiting because of some other lock on one server.
A very similar solution would be:
Read from Random Server - Write to Server A (which replicates automaticly to all others)
This would be more expensive because Standard licence is needed and the first one could be done with express, but would it be better in matters of performance? How fast are those automated replicates? Does all this make sense at all or is there a better nicer way?