Everyone likes the idea of high availability for virtual machines running on Hyper-V, but the price is always a concern.We recently researched this for a client and I thought I'd pass along high level estimates of the costs involved.
- 2 x Dell PowerEdge 510, dual processor, 48 GB RAM, quad port nic, 2 small mirrored drives, Windows Server 2008 Enterprise @ $6000
- 1 x Dell MD3200i (iSCSI) SAN with 8 x 600 GB 15K drives @ $15000
- 1 x 24-port 1Gbps switch for SAN network @ $400
The question becomes....Is this worth it? A single PowerEdge 510 with disk equivalent to the MD3200i listed above is about $9000. That is a lot less money overall.
Most smaller organizations don't have difficulty getting a maintenance window to apply updates on the virtualization hosts. So, the ease of maintenance where you can failover VMs to another host when updating the host operating system is of little benefit. Convenient, but certainly not important for most small organizations.
The server hardware components most likely to fail are disks and power supplies. These component are already redundant on any server that we sell. Consequently the risk of downtime due to hardware failure is already quite low. However, there is a small chance that other critical components such as a motherboard could fail and take down all of virtual machines. Each business needs to decide what the cost of that downtime is and whether they are willing to risk it. One way to mitigate the cost of downtime is increasing warranty service. For example, instead of next day service increase the warranty to 4 hour service. Or for the truly paranoid a second server with just hardware and less drives would be about $4000. You could keep that hardware as spare parts for quick recovery at a much lower cost that a full high availability
For right now, our clients have been willing to take the risk of outages for the cost savings. However, I should note that having a functional disaster recovery plan is an important part of this decision process. That plan identifies realistic time frames for recovery ins various scenarios and allows you to make intelligent decisions.