Overview of Microsoft Cluster Service

--------------------------------------------------------------------------------
Describing Cluster Service and the benefits of using a cluster.


--------------------------------------------------------------------------------
--------------------------------------------------------------------------------

What Is a Cluster?
A cluster is a group of two or more individual servers, called nodes, that work together and represent themselves to the network as a single server. This includes providing access to resources such as file shares, printers, and other network services. The servers in a cluster are both physically and programmatically connected, so they can coordinate communication in response to client requests. Each server in a cluster is able to provide redundant operations in the event of a hardware or application failure. As a result, the cluster provides a high degree of availability for applications and resources.

In addition, a cluster might also be used to implement static load balancing to share the workload among the individual servers in the cluster. If you need dynamic load balancing, you should consider using NLB.

Benefits of Using Cluster Service
Implementing Cluster Service provides a number of benefits. These include the following:

Increased Scalability

Servers can be expanded to provide additional memory and processing resources without interrupting network services.
Cluster Service allows integration of new server resources (both hardware and operating system software) with existing legacy resources. For example, using Cluster Service, you can combine Windows 2000 Advanced Server computers and Windows NT 4 Server computers in the same cluster.

High Availability

Cluster Service can detect a hardware or software failure and quickly assign the application or service responsibility to another node in the cluster. This process is called failover.
Cluster Service allows nodes that come back online after a failure to automatically resume their original responsibilities. This process is called failback.
Cluster Service allows access to network resources and services during planned downtime. For example, if you need to upgrade a node in the cluster, you can manually move the network services that the node is providing to another node. Client requests will not be interrupted while the original node is offline.
Cluster Service reduces single points of failure on your network. While Cluster Service alone cannot guarantee server uptime, it does provide a higher level of availability.
Applications can be distributed over more than one server, which results in a faster response time.

Improved Manageability

Cluster Service allows you to manage the devices and resources in a cluster as if they were all on a single server.
Cluster Service provides a single point of control for administrators. Administration can also be accomplished remotely.
Cluster Service allows applications and services, called resources, to be taken offline for upgrades or maintenance.

Clustering Terminology
The terminology used in Microsoft clusters includes the following:

Active/Active A type of cluster implementation in which each node is capable of managing the resource groups specified in the cluster. When one node fails, the other takes control of the resources. Each node is capable of dynamically assuming another's role.
Active/Passive A type of cluster implementation in which specific resource groups are assigned to a specified node as the resource's primary node. If the primary node fails, the resource will fail over. When the failed node returns online, it resumes control of the resource.
Cluster A group of individual servers working together as a single server.
Common resource A resource that is accessible by each node of cluster. These include the shared drive array and the private cluster network.
Dependencies A reliance between two resources that requires both to run within the same resource group.
Domainlet A substitute for a traditional domain with limited features. Because the nodes in a cluster must belong to the same domain, you can optionally configure the servers to be domain controllers. As a result, the amount of overhead on each server will increase. To minimize this, you can implement a domainlet, which provides limited capabilities regarding groups, policies, and authentication. Only the minimum requirements for service account authentication are included. This makes the use of domainlets for the cluster nodes preferable over a traditional domain.
Failback The process of returning control of a resource or group to the node where it was running before the failure occurred.
Failover The process of moving control of a resource or group from one node to another in the event of a failure.
Group, resource group A collection of resources used for the purpose of configuration and management. The terms group and resource group are used interchangeably; they are also sometimes referred to as failover groups. A group typically includes all of the resources required to support a given application, such as a disk drive, an IP address, a network name, appropriate application services, and so on. If resources are dependent, they must exist in the same group. All resources in a group must be online and on the same node in a cluster. A group is managed as a single logical unit.
IsAlive check Resource Monitors use the IsAlive check to do an exhaustive verification of the state of a resource. If the IsAlive check fails, the resource is moved offline and the failover process begins. You can specify the IsAlive polling interval to control how quickly Cluster Service is made aware of a failed resource.
LooksAlive check A simple check performed by the Resource Monitor to verify that the resource is running properly. If the LooksAlive check fails or cannot determine the state of the resource, the more thorough IsAlive check is used. You can specify the LooksAlive polling interval to control how quickly the Cluster Service is made aware of a failed resource.
Node A server in a cluster.
Offline In reference to a resource: unavailable and unable to provide service to the cluster.
Online In reference to a resource: available and providing service to the cluster.
Quorum resource A common resource that is located on a physical disk of the cluster's shared drive array. It stores a synchronized version of the cluster database. This database contains cluster management data. As a result, the quorum resource must be present in order for node operations to occur. The quorum resource also includes recovery logs that contain the transactions or changes that have been made against the cluster database.
Resources Physical or logical entities managed by Cluster Service, such as file shares, shared printers, and applications. Resources are the basic units managed by Cluster Service and can run on only a single node in the cluster at a time. Resources include both hardware and software components in the cluster. Resources can be brought online or offline and are controlled by only one node at a time.
Virtual server A mechanism by which applications and resources in a cluster can be exposed to clients. A cluster can support more than one virtual server based on how you group the resources and applications in the cluster. (See Figure 1.1.) This allows clients to access the resource regardless of which node is controlling the resource, through the same IP address or NetBIOS name. Each virtual server in a cluster has its own unique IP address and NetBIOS name.



Summary
Cluster Service allows multiple individual servers to share resources and represent themselves to the network as a single server. As a result, a cluster provides hardware redundancy and high availability for network services such as file shares and mission-critical applications. The benefits of implementing a cluster include increased availability, improved scalability, and greater manageability.