A Windows cluster is a collection of one or more Windows servers that work together. MSCS provides uninterrupted service by detecting the failure of applications and servers and automatically migrating resources and workload to other servers in MSCS. Each Windows server that is part of the cluster is called a cluster node. Nodes share one or more SCSI disks, which are called shared cluster disks.
A server cluster is a group of independent servers running the cluster service. MSCS supports a minimum of two nodes and a maximum of eight nodes in a cluster setup. If a node in the cluster is unavailable because of failure or a planned outage, resources and applications move from the failed node to another node in the cluster. The cluster service ensures uninterrupted resource availability to users.
A cluster resource is a basic system entity, such as a physical disk, process, service, network address, or network name. A cluster group is a collection of cluster resources related either logically or physically by dependencies.
When a node provides resources and executes processes belonging to a cluster group, the node is said to “own” the cluster group and all of its resources. A node in a cluster may own different shared cluster disks at different times. However, each shared physical disk and all its logical disks are owned by only one cluster node at a time.
A failover occurs when one or more cluster resources in a cluster group fail and MSCS migrates the troubled group from one node to another. Ownership of the group is then transferred to the new node.
When installing MSCS, you specify the network name of the cluster. This name becomes the name of the virtual servers you install on the cluster. A virtual server is an application server within a cluster group. Executable and other dependent files of a virtual server are installed on a shared cluster disk so that each node in MSCS can execute the programs. Clients connect to the virtual server through a cluster resource network name. The network name provides transparent access to the virtual server regardless of which cluster node currently owns the group.
If a failover takes place, ownership of the group, including the network name, is transferred to a new cluster node. The node restarts all member processes of the virtual servers in the group. Because client computers use the network name to connect to the virtual server, it does not matter that a different node now owns the group.
MSCS uses an intelligent middleware that resides between the operating system and the applications available to users.
MSCS consists of the following components:
Clustering software: Clustering software enables the servers in a cluster to communicate with each other and schedule the allocation of resources between servers. Clustering software further consists of the following components:
Cluster Service: The Cluster Service runs on each cluster server and manages communication between cluster servers.
Resource Monitor: The Resource Monitor manages communication between the Cluster Service and server resources.
Cluster Administrator: Cluster Administrator allows users to manage and configure cluster servers. The Cluster Administrator provides several setup wizards for installing applications and services on the cluster.
The following table lists some of the commonly used terms in a clustering environment:
|
Term |
Definition |
|---|---|
|
Node |
A server that is a member of a cluster. |
|
Resource |
A hardware or software component such as a printer, an IP address, or a disk that is available in a cluster. |
|
Failover or Failback |
The process of moving resources from a failed server to an available server in a cluster. |
|
Group |
A combination of resources that are managed as a unit of failover. The combination is also called Resource Group or Failover Group. |
|
Active / Active |
A state in which multiple nodes in a cluster contain resources or applications that can actively service clients. |
|
Active / Passive |
A state in which one node in a cluster actively services clients while the other node is idle. |
|
Heartbeat |
A repetitive communication sent back and forth between the cluster nodes at specific intervals. If the signal is not returned, the failover process begins. |