A. By multi-site clustering, I mean multiple locations, each of which is on a different IP subnet. Even with Windows 2003, you can create a cluster that spans multiple physical locations, but you have to use a Virtual LAN (VLAN), so the network appears to be a single IP subnet. A VLAN is typically a very expensive solution.

This single subnet requirement is because a cluster can have multiple IP addresses, but they’re in an AND logic, which means they all have to be available for a resource to be brought online. This won’t work if you want IP addresses in different subnets, because your entire goal was to have different addresses for the different subnets in the various locations.

With Server 2008 and above, you can use an OR logic, which means you can have different IP addresses in different subnets, and the IP address that can make connectivity will come online and enable the dependent resources. So if the resource is hosted in location A, the IP address for the location A subnet will come online and enable dependent resources such as the network name. If the resource fails over to location B, the IP address for location B's subnet will come online. This ability to have different IP addresses in different subnets is primarily what allows you to stretch a failover cluster across different IP subnets in Server 2008.

As you span different locations, you’ll most likely get slower and higher latency links and heartbeat communication may be delayed or lost, which could cause false failovers. Windows Server 2008 allows a great deal of granularity in configuring the heartbeat’s acceptable limits to take into account slower, higher latency links to avoid the false failovers.

Related Reading:

Check out hundreds more useful Q&As like this in John Savill's FAQ for Windows. Also, watch instructional videos made by John at ITTV.net.