Server Cluster Overview

Applies To: Windows Server 2003, Windows Server 2003 R2, Windows Server 2003 with SP1, Windows Server 2003 with SP2

This section gives a general overview of server clusters and the Exchange server cluster at Microsoft that supports messaging within the company.

The following terminology is used in this section and throughout the document.

• Active node. The cluster server that currently owns cluster group resources and responds to network requests made to those services.

• Alternative passive node. The cluster server that does not currently own cluster group resources but is available if the active fails over and the primary passive node is unavailable.

• Primary passive node. The cluster server that does not currently own cluster group resources but is available if the active node fails over.

• Virtual server. A collection of services that appear to clients as a physical Windows-based server but are not associated with a specific server. All virtual servers must include a Network Name resource and an IP Address resource.

Server clusters can take two forms: active/passive clusters and active/active clusters. In active/passive clustering, the cluster includes active nodes and passive nodes. The passive nodes are only used if an active node fails. In active/active clusters, all nodes are active. In the event of a failover, the remaining active node takes on the additional processing operations, which causes a reduction in the overall performance of the cluster. Active/passive cluster configurations are generally recommended over active/active configurations because they often increase performance, availability, and scalability.

The Exchange server cluster is based on an active/passive model, with a few alterations that maximize the performance of the hardware. The Operations and Technology Group (OTG) has configured a multinode cluster that consists of four active nodes, one primary passive node, and two alternate passive nodes. The four active nodes provide most of the processing operations for the messaging environment. Like most active/passive configurations, the primary passive node is used only when one of the active nodes fails or to support the process of rolling upgrades. The two alternate passive nodes are only used for messaging if failover occurs and the primary passive node is not available. To take full advantage of the available hardware, the alternate passive nodes are also used to facilitate the backup and upgrade processes. This unique configuration supports a total of 16,000 mailboxes.

The following topics are described in this section:

• Diagram 1. Exchange Server Cluster

• Reasons for Using Alternative Passive Nodes

• Diagram 2. Failover Process of the Server Cluster

• How OTG Configures the Exchange Server Cluster

Diagram 1 Exchange Server Cluster

Diagram 1 shows the four active nodes, the primary passive node, and the two alternate passive nodes that make up the Exchange server cluster. All of the nodes are connected to two storage area networks (SANs). The two SANs provide communication between the nodes and the shared storage components, which are two redundant Enterprise Virtual Arrays. For more information about how storage is configured in this cluster, see Storage Components later in this document.

Each node in the cluster has two adapters that support a public network and a private network. This is a best practice that is described later in this document. The private network is reserved exclusively for internal communication between cluster nodes. The public network supports both node-to-node and client-to-cluster communication.

Reasons for Using Alternative Passive Nodes

The four active nodes and the primary passive node each have four hyperthreading 1.9-gigahertz (GHz) processors and 4 gigabytes (GB) of RAM. The two alternate passive nodes are less expensive servers with dual 2.4-GHz processors and 2 GB of RAM.

Although it is rare, the alternate passive nodes can be used to support Exchange processing if an active node fails and the primary passive node is unavailable. Instead, the alternate passive nodes serve two main functions. First, the alternate passive nodes are primarily used to facilitate stage 2 of the backup process. Second, these nodes help during scheduled rolling upgrades and/or patch application to the operating system and/or the host.

Diagram 2 Failover Process of the Server Cluster

Diagram 2 shows how the primary passive nodes and alternate passive nodes are used during failover. The four active nodes provide the primary processing operations for the messaging environment. The primary passive node is used when one of the active nodes fails. The two alternate passive nodes are used for messaging only if failover occurs and the primary passive node is not available.