In this ZDNet article, they provide an overview of virtual server host clustering and the prerequisites that must be met for it to work. The article assumes that the reader is familiar with the Cluster Service technology included with Windows Server 2003.  Clustering is described as a technology used when you want to increase the availability of a server, service or application so it doesn't become a single point of failure. As an example, if you have a critical server running on your network, it can be placed in a cluster with another server that is standing by to take over in the event of failure.  Windows Server 2003 Enterprise and Datacenter Editions both include a technology known as Cluster Service that is designed to increase the availability of your critical components.  By combining this Cluster Service technology included with Virtual Server 2005 R2 (VS2005 R2), you can even increase the availability of your guest OSs running on a host.

What is host clustering?
With virtual server host clustering, you combine the clustering technology in Windows Server 2003 with VS2005 R2. The hosts, running VS2005 R2 on Windows Server 2003, are set up in a cluster configuration. A minimum of two hosts are required. The number of hosts that can be supported in a single cluster will depend on which version of Windows Server 2003 is installed.

The guest OSs are installed on a host in the usual fashion. Should a node in the cluster fail, one of the other nodes in the cluster will automatically take over support of the guest OSs. 

Benefits
One of the obvious benefits to host clustering is that it increases the availability of a guest OS, should the host on which it is running fail. Along with this, clustering also provides additional benefits:

• It allows for scheduled downtime for maintenance with minimal interruption. Guest OSs can be run on another node in the cluster while the original host is taken offline for maintenance.
• Applying software updates can be applied to a host without having to take the guest OSs offline. Again, the guest OSs can be moved to another physical server while software updates are applied to the original host.
• It allows you to cluster OSs and applications that are not cluster-aware. Such components are clustered within the context of a guest OS. Availability of the guests is increased, regardless of the OS and applications running within them.

Check out the entire article for more information, here.