What Is A Cluster Network?

A cluster consists of several servers interconnected by an input / output system. These servers are connected to storage media and are managed by distributed resource management software (DRM). Its components are in continuous improvement: blade servers, fast InfiniBand I / O technology and more complex DRM software. These parts combine to make the cluster a useful tool for IT managers.

A cluster consists of several servers interconnected by an input / output system. These servers are connected to storage media and are managed by distributed resource management software (DRM). Its components are in continuous improvement: blade servers, fast InfiniBand I / O technology and more complex DRM software. These parts combine to make the cluster a useful tool for IT managers.

Chinese name

Server cluster

Foreign name

Server Cluster

Make up

Several servers connected to the input / output system

Cluster concept

Network load balancing

Cluster algorithm

Fully distributed

Main application of server cluster

The main application of server cluster technology is the function of network load balancing. Network load balancing uses a cluster of two or more host computers working together to provide servers with high availability and scalability. Internet clients use one IP address or a group of addresses to access the cluster. The client cannot distinguish between a cluster and a single server. Server applications do not indicate that they are running on a cluster. However, a network load balancing cluster is very different from a single host running a single server application because it can provide uninterrupted service even in the event of a cluster host failure. The cluster also responds faster to client requests than a single host.

If a host fails or goes offline, Network Load Balancing brings high availability by redirecting incoming network traffic to a working cluster host. Existing connections to the offline host will be lost, but Internet services will still be available. In most cases (for example, in the case of a web server), the client software can automatically retry failed connections, and the client receives a response with a few seconds of delay. Network Load Balancing brings variable performance by distributing incoming network traffic among one or more virtual IP addresses (cluster IP addresses) assigned to a Network Load Balancing cluster. Then, the hosts in the cluster respond to different client requests or even multiple requests from the same client at the same time. For example, a web browser can get multiple images from all single web pages from different hosts in a network load balancing cluster. This improves processing speed and reduces response time to clients.

Network load balancing enables all cluster hosts on a single subnet to detect incoming network traffic to the cluster IP address simultaneously. On each cluster host, the network load balancing driver acts as a filter between the cluster adapter driver and the TCP / IP stack in order to distribute traffic between hosts.

Server cluster clustering algorithm

Network load balancing uses a completely distributed algorithm that statistically maps incoming client IP addresses and ports to cluster hosts. This process takes place without any communication between the hosts. When an arriving packet is found, all hosts perform this mapping at the same time to quickly determine which host should process the package. This mapping remains the same until the number of cluster hosts changes. Compared with centralized load balancing applications, network load balancing filtering algorithms process packets more efficiently because the former must modify and retransmit the packets.

Network load balancing controls the distribution of TCP and UDP communications from Internet clients to selected hosts in the cluster in the following ways: After network load balancing is configured, all hosts in the cluster receive incoming clients that are passed to the cluster IP address request. Network load balancing filters incoming datagrams that pass to the specified TCP and UDP ports before these datagrams reach the TCP / IP protocol software. Network Load Balancing manages the TCP and UDP protocols within TCP / IP, controlling their operation on a port-by-port basis.

In multicast mode, Network Load Balancing can provide Internet Group Management Protocol (IGMP) support to limit exchange flows. Except for TCP and UDP communication on a specified port and IGMP communication in multicast mode, network load balancing does not control any incoming IP communication. It does not filter other IP protocols (for example, ICMP or ARP), except as described below. Note that when using a cluster IP address, you should see duplicate responses from specific point-to-point TCP / IP applications, such as ping. If needed, these applications can use a dedicated IP address for each host to avoid this.

To coordinate its operation, Network Load Balancing hosts periodically exchange heartbeats within the cluster (see Internet Group Management Protocol (IGMP) for more information). IP multicasting allows hosts to monitor cluster status. When the state of the cluster changes (such as when a host fails, leaves, or joins the cluster), Network Load Balancing invokes a process called "aggregation," in which the hosts exchange a limited number of messages to determine the cluster's new Consistent status and assign the highest host priority to the host, that is, the new default host. When all cluster hosts are consistent in the correct new cluster state, they will record the completion of the aggregation in the Windows event log. It usually takes less than 10 seconds to complete this process.

During the aggregation process, the remaining hosts continue to process incoming network traffic. Client requests to the working host are not affected. When the aggregation is complete, the traffic targeted to the failed host is redistributed to the host that is still working. Load-balanced traffic will be re-divided among hosts that are still working to achieve the best possible new load balancing for specific TCP or UDP ports.

If a host is added to the cluster, aggregation allows the host to receive its own load-balanced communication. The expansion of the cluster does not affect ongoing cluster operations, and its implementation is transparent to Internet client and server applications. However, when "Client Similarity" is selected, it may affect client sessions across multiple TCP connections because clients may be remapped to different cluster hosts between connections. For more information on similarities, see Network Load Balancing and State-Controlled Connections.

Network load balancing assumes that the host is working properly in the cluster for as long as it exchanges heartbeats with other cluster hosts. If, during multiple heartbeat handshake, no other host receives a response from any member, they will initiate aggregation and redistribute the load that should have been handled by the failed host.

Server cluster cluster function

Server Cluster Features in WIN 2003

The server cluster feature was originally designed for the Microsoft Windows NT® Server 4.0 operating system, and this feature has been significantly improved in the Microsoft Windows Server 2003 Enterprise Edition and Windows Server 2003 Datacenter Edition operating systems. With server clustering, you can connect multiple servers together to provide high availability and manageability for the data and programs running in the cluster. Server clustering provides three main clustering technology benefits:

Higher availability. Allows services and applications in a server cluster to provide uninterrupted service in the event of a hardware or software component failure or during scheduled maintenance.

Higher scalability. Support for adding multiple processors (up to 8 in Windows Server 2003 Enterprise Edition and up to 32 in Windows Server 2003 Datacenter Edition) and additional memory (in Enterprise Edition, Random Access Memory [RAM] Up to 8 GB, and up to 64 GB in Windows Server 2003 Datacenter Edition) to expand the server.

Higher manageability. Allows administrators to manage devices and resources across the cluster as if they were a single computer.

The cluster service is one of two complementary Windows cluster technologies (provided to extend the Windows Server 2003 and Windows 2000 base operating systems). Another clustering technology is Network Load Balancing (NLB). This technology is complementary to server clusters and can support front-end applications and services such as Internet or intranet sites, web-based applications, media streaming, and Microsoft Terminal Services to support highly available and scalable clusters.

The emergence and use of computer clusters has a history of more than ten years. As one of the earliest cluster technology designers, G. Pfister's definition of a cluster is "a parallel or distributed system consisting of a collection of fully interconnected computers that can be used as a unified computing resource."

By combining several server computers into a unified cluster, multiple servers can share the computing load without the user or administrator having to understand the details. For example, if any resource in a server cluster fails, regardless of whether the failed component is a hardware or software resource, the cluster as a whole can continue to provide services to users using resources on other servers in the cluster.

In other words, when a resource fails, users connected to the server cluster may experience a short period of performance degradation, but will not completely lose access to the service. When more processing power is needed, administrators can add new resources through a rolling upgrade process. During this process, the cluster will remain online as a whole, and it will not only be available to users, but its performance will be improved after the upgrade.

The Windows Server 2003 Enterprise Edition and Windows Server 2003 Datacenter Edition operating systems are designed and developed to meet the requirements of users and businesses for cluster technology. The main goal is to develop an operating system service that meets the clustering needs of most businesses and organizations, not just small and specific market segments.

Microsoft market research shows that as small and medium-sized business organizations increasingly rely on databases and e-mail for their daily operations, they have a high demand for highly available systems, and this demand is growing. Ease of installation and management is considered the most critical requirement for organizations of this size. Microsoft's survey also shows that large enterprises with high requirements for high performance and high availability are also increasingly interested in Windows-based servers.

The server cluster service developed as an integrated extension of the Windows NT, Windows 2000, and Windows Server 2003 base operating systems is derived from this market survey. This service is consistent with its design goals, through which multiple servers and data storage components can be connected into one easy-to-manage unit, a server cluster. For systems running Windows Server 2003 and Windows 2000-based applications in large and small businesses, server clustering capabilities will give them high availability and manageability. The server cluster feature also provides the application programming interfaces and tools necessary to develop new cluster-aware applications that take advantage of the highly available capabilities of server clusters.

Server cluster installation configuration

The following describes how to install and configure the cluster service on a Windows Server 2003 system:

Step 1. Click "Start" "All Programs" "Administrative Tools" "Cluster Administrator" menu in turn to open the "Cluster Administrator" window and automatically open the "Open Connection to Cluster" dialog. This dialog box provides users with three options for creating a new cluster, adding a node to the cluster, and opening a connection to the cluster. Select the "Create a new cluster" command in the "Action" drop-down menu and click the "OK" button.

Step 2. Open the New Server Cluster Wizard and click the Next button on the Welcome Wizard page. Enter a cluster name in the Cluster Name and Domain wizard page that opens, and the name must be a valid computer name. You also need to choose a domain name, and the new cluster will be created in this domain. Use the default domain name and click the "Next" button.

Step 3. In the "Select Computer" wizard page that opens, prompt the user for the first node computer where the new cluster will be installed. By default, the local server will be selected automatically. Click the "Next" button.

After you click the Advanced button and select the Advanced (minimum) configuration check box, you need to manually add shared devices after the New Server Cluster Wizard completes the final steps.

Step 4, followed by the New Server Cluster Wizard will use the information provided by the user to analyze all the factors necessary to successfully create a cluster. The wizard mainly analyzes through the following five aspects:

(1) Existing cluster: verify that no cluster with the same name exists on the network;

(2) Establish a node connection: connect to the target server and initiate the cluster creation process;

(3) Check node feasibility: determine whether the target server meets all prerequisites;

(4) Find public resources on the node: determine the availability of shared resources, including enumerating shared disk resources (to create a quorum) and network adapters (to install a cluster network);

(5) Check the feasibility of the cluster: verify that the creation of the cluster is possible and specify resources for quorum.

The wizard will use a tick to indicate success, a cross to indicate failure, and a yellow triangle with an exclamation point to indicate a warning, where the warning sign can be ignored. If there is a problem in any aspect of the analysis, you must clear the fault before proceeding. After passing the analysis verification, click the "Next" button.

Step 5. Open the "IP Address" wizard page. At this time, you need to set the IP address of the cluster. This address will be used as the destination address for DNS resolution of the cluster name, and the address should be on the same network segment as the IP address of the "public connection". In this example, enter a free IP address 10.115.1.168 on this network segment. After setting, click the "Next" button.

Step 6. In the Cluster Service Account wizard page that opens, you need to provide a domain user account to manage the cluster. This account will be placed in the local Administrators group and given the appropriate permissions on the target server. Enter the cluster user account Cluster and its password created earlier, and click the "Next" button.

Step 7. Open the Suggested Cluster Configuration wizard page, which lists all the information that the user has set up so far. After confirming that it is correct, click the "Next" button.

In step 8, the wizard begins to create and start the server cluster. This process is divided into 4 main parts:

(1) Reanalyze the cluster: Repeat the steps originally selected when verifying the feasibility of the cluster. Once this feasibility has been verified, the cluster configuration has been initialized;

(2) Configure the cluster service: Assign the cluster service account to the appropriate group and give it the appropriate user rights. Then create and start cluster-related services, create and configure the cluster database, and create and start the Cluster Service;

(3) Configure the resource type: configure the Generic Script and Majority Node Set resource types;

(4) Configure resources: Create, configure, and start resources (including quorum resources) in the cluster group.

Click "Next" button after creation.

After the cluster service is installed, it automatically connects to the cluster.

If you need to cancel the cluster configuration because you cannot connect to the cluster due to a configuration error, you can enter the command line "cluster node node name / forcecleanup" in the Command Prompt window and press Enter.