The clustering-solution software included with Microsoft Windows NT Server, Enterprise Edition. Microsoft Cluster Server (MSCS) supports clusters of two servers that are running Windows NT Server, Enterprise Edition. It provides network administrators with a scalable, centralized solution for networks that require high availability of resources. When network clients try to connect to shared resources on MSCS, they do so in the same way that they connect to any other network server. Users can thus learn to work with clusters without requiring additional training.
Clustering also provides static load balancing - that is, the distribution of processes across servers. This is different from fault tolerance, which provides error-free, nonstop availability of resources, usually through a backup of the primary system that remains idle until a failure occurs. In a cluster, the redundant systems are active, not passive.
With MSCS, when one server (node) in a cluster fails or is taken offline, the other server takes over the failed server’s operations. Clients accessing resources on MSCS experience little or no interruption in their work because MSCS automatically moves the resource functions from one cluster node to the other.
MSCS nodes in a cluster are connected using a network, which typically is a Small Computer System Interface (SCSI) bus. MSCS provides users with a single, virtual image of the cluster. If one node in the cluster fails, the applications on the failed node are made available on the other node. Client access to network applications continues with little or no interruption. In most cases, if the interruption is detected in 5 seconds, services will be available again in 30 seconds or less, depending on how long it takes to restart the application.
For example, suppose one node in a cluster runs Internet Information Services (IIS) to service Web clients, while the other node runs a Microsoft SQL Server database. If the first node fails, the resource (IIS) fails over to the second node, which now responds to both Web and SQL requests from clients.
MSCS has two components:
In MSCS, a resource (such as a network application or shared resource) can be hosted on only one node at a time. If you configure the resource to fail over, the operation of the resource moves from the failed node to the working node in response to a failure. Related or dependent resources can also be organized into resource groups, which can also run on only one node at a time. If a resource group has an IP address and network name assigned to it, the group’s resources appear to network clients as individual servers called virtual servers. Users can access applications or services on a virtual server exactly as they would if the application or service resided on a physical server.
MSCS clusters can work with only one interconnect, but two interconnects are strongly recommended to eliminate a single point of failure for the cluster.
For information on clustering in Windows 2000, see the entry Windows Clustering elsewhere in this work.