Failover Clustering Architecture

SQL Server Setup Help

SQL Server Setup Help

Failover Clustering Architecture

Microsoft® SQL Server™ 2000 failover clustering increases server availability by allowing a system to automatically switch the processing for an instance of SQL Server from a failed server to a working server. For example, an instance of SQL Server can quickly restore database services to a Web site or enterprise network even if the server running the instance fails. SQL Server 2000 implements failover clustering based on the failover clustering features of the Microsoft Clustering Service (MSCS) in Windows NT® 4.0 and Windows® 2000.

The type of MSCS failover cluster used by SQL Server 2000 consists of multiple server computers (two on Windows NT 4.0, up to four on Windows 2000 Datacenter Server) that share a common set of cluster resources, such as disk drives. Each server in the cluster is called a node. Each server, or node, is connected to the network, and each node can communicate with each other node. Each node runs the same version of MSCS.

The shared resources in the failover cluster are collected into cluster groups. For example, if a failover cluster has four clustered disk drives, two of the drives can be collected in one cluster group and the other two in a second cluster group. Each cluster group is owned by one of the nodes in the failover cluster, although the ownership can be transferred between nodes.

Applications can be installed on the nodes in the failover cluster. These applications are typically server applications or distributed COM objects that users access through network connections. The application executables and other resources are typically stored in one or more of the cluster groups owned by the node. Each node can have multiple applications installed on it.

The failover cluster nodes periodically send each other network messages called heartbeat messages. If the MSCS software detects the loss of a heartbeat signal from one of the nodes in the cluster, it treats the server as a failed server. MSCS then automatically transfers the cluster groups and application resources of that node to the other nodes in the network. The cluster administrator specifies the alternate nodes to which cluster groups are transferred when any given node fails. The other nodes then continue processing user network requests for the applications transferred from the failed server.

For more information about MSCS, see the Windows NT Server, Windows 2000 Server, Windows 2000 Advanced Server, or Windows 2000 Datacenter documentation.