JSI Tip 6544. The Microsoft Windows Cluster Service (MSCS) clustering model.

MORE INFORMATION

These are two different types of cluster servers, the fail-over cluster and the load-balancing cluster. The following two different types of fail-over clusters are available:

• The shared-everything cluster.
• The shared-nothing cluster.

The MSCS Cluster service supports the shared-nothing fail-over style of server cluster infrastructure.

The Shared-Nothing Cluster Model

In the shared-nothing cluster model, all servers or cluster nodes in the cluster can own the shared cluster resources. However, only one node may own and manage these resources at any one time. This means that while cluster node A owns a physical disk, no other cluster node can access it. Node A has exclusive ownership of this shared disk until it is either manually moved to another cluster node, or until node A fails and another cluster node assumes ownership of the resource.

Explanation

If you configure an Active/Active server cluster, multiple instances of a service or program can exist at the same time, but they must each have their own individual databases. For example, if you install Microsoft SQL Server Enterprise Edition on a Windows 2000-based server cluster, you can install two instances of SQL Server. For example, SQL-1 and SQL-2. Each instance of SQL Server must have its own unique database to manage. In this case, while cluster node A owns the SQL-1 resource, cluster node B cannot start its SQL-1 instance, and cannot access the SQL-1 instance database.

There will never be more than one cluster node that is running an instance of SQL-1. The other cluster nodes that can own the SQL-1 resource must wait for the node that owns it to fail so one of them can host the resource. In a shared-nothing Active/Active configuration, both nodes can host resources at the same time, but they cannot host the same resources at the same time.

The Shared Everything Cluster Model

In the shared-everything cluster model, all servers or cluster nodes in the cluster are given equal access to the shared resources. This is done by using a Distributed Lock Manager (DLM). The DLM monitors all the shared cluster resources and allows the nodes to access the resources when they are available.

Explanation

In this model, if cluster node A wants to access the cluster resource disk X, it requests access to disk X from the DLM. If disk X is not being used, the DLM grants a lock on disk X to cluster node A. This node can then perform its required tasks with disk X. If cluster node B must access disk X while node A already has a lock on it, cluster node B is placed in a queue by the DLM for access to the cluster resource when cluster node A is finished with it.

Load Balancing Cluster Model

This is known as the Network Load Balancing (NLB) service in Windows 2000 and Windows Load Balancing Service (WLBS) in Microsoft Windows NT 4.0.

For additional information about how to configure network load balancing, click the following article number to view the article in the Microsoft Knowledge Base:

240997 Configuring Network Load Balancing