Server cluster

Server cluster means will many server centralized, together with the same kind of services in the client looks like that is only a server cluster can use more computer for parallel computing to obtain high calculation speed, also can use more computer do backup, making any machine broke down the whole system can still normal operation

Server cluster

Once on the server cluster service installed and running, the server can join cluster. Cluster of operation can reduce single-point fault the amount, and realized the resources of the swarm of high availability. The following sections briefly introduces clustered create and clustered operation of the node behavior. Note: the information about installation clustered servers, please refer to the Windows server 2003 product family help and deploy guidelines.

Create clustered

In the server cluster products containing used in server installed software and create a new cluster swarm of swarm installation practical tools. Create a new cluster, the first in selection as a cluster of the first members of the practical tools to run on a computer. The first step is to confirm whether cluster name and create clustered database and initial cluster member list to define new cluster. Windows server 2003 clustered adds a cluster management setup wizard and using cluster. Exe command-line interface to create (including from remote create) clustered function.

The second step is to create clustered, add all the members using cluster for Shared data storage devices. So, create new clustered with a node, their local data storage equipment and clustered Shared resources - often a disk or data storage and connection medium resources.

The last step is to create clustered in another will be clustered, members of the run on every computer installed practical tools. When will the new node added to cluster when new nodes will automatically get the original members from clustered copies of existing cluster database. When the node join or form cluster, cluster service will update this node private configuration database copy.

Form cluster

If the server cluster service running up the other and cannot find clustered node, it can form a cluster. To form a cluster, node must be able to get the exclusive rights to the arbitration resources.

After the initial swarm, formed the first node cluster will include clustered configuration database. Whenever there is a new nodes, the new node crypt_e_hash_value will be in local gain and sustain clustered configuration database copy. Arbitration resources by a recovery log (with node containing the clustered configuration and independent state data) form of the latest version of storage configuration database.

In the cluster operation, the cluster service use arbitration restored log perform the following operations:

To ensure that only a set of activities, mutual communication node can form cluster

Only when a node can obtain the control of arbitration resources will allow it to form newly-constructed swarm

A node can only when the node with control arbitration resources allowed communication, it join or stay current in the cluster

The other nodes and from clustered cluster service management interface perspective, when the swarm, cluster formed each node may be in three different state of a kind. An event handler will record these state, while event log manager will these state copied to the clustered other nodes. Cluster service states including:

Offline. At this time of node is not fully effective cluster members. The node and its clustered servers in operation, may also could not run.

Online. At this time of node is completely effective cluster members. It teaches clustered updates to the database, to the arbitration algorithm exert their influence, maintenance, and can have heart communications and operation resources group.

Pause. At this time of node is completely effective cluster members. It teaches clustered updates to the database, to the arbitration algorithm exert their influence, maintenance heart communications, but it can't accept resource group. It can only support it current have those resources group. The state, is to provide suspended allow execution some maintenance. Most of the server cluster components will online and suspended as equivalent condition.

advantage

A, cluster system can solve all the server hardware failure, but when one server any fault, such as: hard drive, memory and CPU, motherboard, I/O board and power failure, run on this server on the application will switch to other server.

Second, cluster system can solve software system problems, we know, in computer system, the users use applications and data, and the use of system operating above the operating system, operating system and run on servers. So, as long as the application system, operating system, server any of the three fault occurs, the system actually stopped providing services to the client, such as our common software crash, is one of this kind of circumstance, although the server hardware intact, but the server still can provide service to the client. But the biggest advantage of cluster server is the fault of the monitoring is based on application, that is, as long as the server application stop running, other relevant server will take over the application, and don't pay attention to the application for what reason stop running.

Three, cluster system can solve the application system caused by human error to stop work, for example, when administrators to one server operating errors led to this server downtime, therefore run in this server application system also stopped running. Due to the cluster is monitored for application, so the other relevant server will take over the application.

faults

We know only in a application of cluster server, if this application is out of order, the other one server will restart the application, and took over the data in Shared disk lockers, then make application area to work properly. We know that the whole application takeover process generally requires three steps: detect and faults are identified, backup server restart the application, takeover Shared data section. Therefore in the process to switch, in principle, it can take time according to the applied size different switching time also will be different, the greater the application switching time longer.