I was wondering why exactly I need CSV for failover-clustering. Would be happy if it was possible (with SCVMM for example) to have a cheap-man's-failover without shared storage:
- (Continually doing) asynchronous Hyper-V replication between servers
- Live migration (planned failovers)
- Failover to the replication target (unplanned failovers)
- Orchestrated by the SCVMM server

That'd be good enough for most shops and be way cheaper and easier to manage than a cluster with proper, reliable shared storage. An alternative would be a useable, integrated software-based storage replication solution.

Will that ever come?