Design for a Failover Cluster with Multiple Services and Applications
Applies To: Windows Server 2008
This design starts with the basic idea of clustering a service or application and expands it to include multiple services and applications on one cluster. For additional designs, such as a simpler design for a clustered file or print server, see Mapping Your Deployment Goals to a Failover Cluster Design.
Note that for the maximum availability of any server, it is important to follow best practices for server management—for example, carefully managing the physical environment of the servers, testing software changes before placing them into production, and carefully keeping track of software updates and configuration changes on all clustered servers.
With any design that involves multiple services and applications on a cluster, it is crucial to consider server loads and server capacity. Any given node on the cluster must have enough capacity to perform reasonably well with the load it will carry in a typical failover situation. It is also important to consider the relative importance of each service or application in relation to your organization’s goals. If one of the services or applications is significantly less important, it might be more appropriate not to run that application as a clustered application. Instead, you might place it on a non-clustered server. This would mean that you could maximize the server capacity devoted to supporting the clustered applications, so that if one of the clustered servers had minor performance problems, the clustered applications would be affected only minimally, or not at all.
Design with clustered applications on a three-node cluster
The following diagram shows a failover cluster design with four clustered applications on a three-node cluster, when all nodes of the cluster are running normally. As indicated by the boxes behind each server, the servers are running at approximately half their working capacity. That is, each server could double its current load and still function adequately.
Four clustered applications on a three-node cluster
The following diagram shows the same design when Node 1 has failed over to Node 2.
Four clustered applications with one application failing over
As can be seen, if one node tried to handle all the applications that exist on the cluster, that node would be overloaded, but two nodes can adequately handle all the applications, assuming the applications are distributed similarly to the diagram. In any case, most three-node clusters will only continue to function if two nodes are up, not just one. This is because clusters with an odd number of nodes normally use the Node Majority quorum option, which would mean that in order to function as a cluster, a majority of the nodes (two) would need to be running at any given time. Therefore, the minimum number of nodes required to keep the cluster functioning (two) is enough to handle the total load of the applications on the cluster.
To see more diagrams showing how failover, failback, and quorum work in this design, see Example, Failover Cluster with Multiple Services and Applications.
For information about configuring this design, see Checklist: Failover Cluster with Multiple Services and Applications (http://go.microsoft.com/fwlink/?LinkId=129125).