Designing Federated Database Servers for High Availability
The data for a large Web site or internal online transaction processing (OLTP) system must be highly reliable. In a clustered application tier, the loss of one server may decrease system performance, but it will not stop the whole system. The remaining servers in the cluster rebalance the load until a replacement server can be plugged into the cluster.
Although SQL Server does not support this kind of load-balanced clustering, it does support Microsoft Cluster Services failover clustering. Failover clustering supports one to four servers per cluster depending on the operating system. The cluster appears to applications as a single virtual server. If the primary server node fails, another node detects the loss of the primary and automatically starts servicing all requests sent to the virtual server. The cluster remains running under the alternative node until the primary server is repaired or replaced. Failover clustering helps provide high availability, but it does not perform any load balancing.