A. Clustering makes two or more machines look like one logical entity on the network. In theory, services that run on the cluster can survive the failure of one (and perhaps more) nodes without any ill effect.
Microsoft's clustering solution, Microsoft Cluster Server (MSCS), is part of Windows NT Server, Enterprise Edition. MSCS lets specially written applications run on two-node clusters. When one node (or a component on it) fails, the second node automatically becomes active and takes over the processing load.
Active/passive clusters run their applications on one node; the other node waits for the first one to fail. Active/active clusters run applications on both nodes at the same time, doubling the effective power of the cluster. Shared-none clusters share no components between the two nodes. Shared-disk clusters share a single disk subsystem between the cluster nodes.
Exchange Server 5.5, Enterprise Edition, can run as an active/passive, shared-disk cluster, which means that the disk subsystem is a single point of failure for the entire cluster and that an extra server sits idle much of the time (but provides additional uptime). Other vendors also offer clustering solutions.