Server cluster system implementation Xiangjie

2010-09-14  来源:本站原创  分类:Internet  人气:147 

First, the basic concept of cluster

There is a common way to substantially increase the server's security, this is the cluster.

Cluster cluster technology can be defined as follows: a group of independent servers in the network performance as a single system, and to model a single system to manage. This single system for the customer workstation to provide highly reliable service.

Most of the model, all the computers in the cluster have a common name, either the cluster services running on the system can be used by all network customers. Cluster must be able to coordinate management of each separate component of the mistakes and failures, and can be transparently added to the Cluster component.

A Cluster contains multiple (at least two sets) have shared data storage server. Any one server running an application, the application data is stored in the shared data space. Each server's operating system and application files stored in their respective local storage space.

Cluster nodes within the server communicate with each other through an internal LAN. When a node server failure, this server application is running on a server in another node is automatically taken over. When a failure of application services, application services will be restarted or another server to take over. When any of the above fault, the customer will be able to quickly connect to the new application services.

Second, the cluster hardware configuration

Dual mirror server

Dual mirror server in the cluster system is the hardware configuration and price of the cheapest most simple solution, often mirroring services, hardware configuration requires two servers, each server in a separate operating system hard drives and data storage drives, each server has connected to the card with the client, while a pair of mirror image function card or complete the card.

Mirror server with a simple configuration, easy to use, low cost many advantages, but mirror image of data required by the network approach, by mirroring data synchronization software, requiring occupation of the network server CPU and memory resources, the mirror server's performance than a single performance of the server is lower.

Some mirror server cluster system uses memory mirroring technology, this technology eliminates the Suoyou of applications and network operating system in both servers Jingxiangtongbu, when the host fails, Bei Fen Ji can be made almost no sense under the circumstances to take over all applications. Because the two server memory exactly the same, but when the system applications with deficiencies which lead to system downtime, the two servers will synchronize downtime. This is also the mirror of memory card or network card for data synchronization, a large amount of data read and write in the course of two servers will have in some state data is not synchronized, the mirror server for those with smaller budgets, less demanding users on the cluster system .

Hardware configuration examples:

  • Two web servers
  • Server operating system hard drive 2
  • Server data storage hard disk, as users need to determine
  • Server mirroring card (part of the software can use standard network card) 2
  • NIC 2 web services

    3, two-plane and the disk array cabinet

    Two-machine system with the mirror server than the two-machine interconnected with the disk array cabinet structure of production more third-party disk array cabinet, at present, Howe companies, fine companies, industry, many companies produce a disk array cabinets, the disk array cabinet equipped with disk array controller, the array can be directly counter to counter in a logical drive array hard disk configuration. Disk array cabinets and server through the SCSI cable connected to ordinary SCSI card, system administrators need to configure the cabinet directly on the disk array.

    Double machine with the disk array cabinet interconnect structure does not use memory mirroring, it requires a certain switching time (usually 60? D? D180 seconds), it can have countryside to avoid the application itself because the defects cause the system to all the downtime, At the same time as all of the data stored in the home all the disk array cabinet, when the work machine fails, the backup machine take over the work machine, read data from the disk array, it will not produce data of sync, as this is kind of program does not require network mirror synchronization, so the performance of the server cluster program structure much higher than the mirror server.

    Double machine with the disk array cabinet interconnect structure among the shortcomings that exist in the system a single point of wrong defects, the so-called single-point error is a part when the system or an application failure, causing all systems are all downtime. In this system, the disk array cabinet will lead to a single point of fault, when the disk array cabinet appear logical or physical failure, all stored data will be lost, therefore, in the selection of such programs, the need to use a quality and after-sales services to better products.

    Hardware configuration examples:

  • Two web servers
  • Server operating system hard drive 2
  • Third-party production of a disk array cabinet
  • SCSI disk cabinet dedicated wire 2
  • Hard disk array storage cabinet, as the user needs the data to determine
  • NIC 2 web services

    In addition, some manufacturers have better technology solutions, such as HP.

    4, HP Dual dual-control Fault Tolerant System

    HP NetServer for the two-plane double-fault-tolerant control system provides high-quality and highly reliable hardware based. .

    HP Dual dual-control system combines the fault-tolerant safety and reliability of HP server technology with the advantages of Cluster, each with two advantages.

    Hardware configuration examples:

  • HP L system, two web servers
  • Server operating system hard drive 2
  • HP hard drive storage cabinet (SS / 6, RS / 8, RS/12) a
  • SCSI disk cabinet dedicated cluster adapter cable 2
  • Hard disk storage cabinets, as the user needs the data to determine
  • HP Cluster-specific array of cards 2
  • NIC 2 web services

    5, HP Fibre Channel Dual dual-control cluster

    Fibre Channel is a connection standard, can be used as an alternative to SCSI solutions, Optical technologies high-bandwidth, Kang Dianciganrao, transmission Juli far, high quality, strong Deng features extended Neng Li, Mu Qian Zhong Cai in the FC-AL loop Shang can access to 126 devices.

    Optical device provides several enhanced connectivity technology, which greatly facilitates the users. Server system can connect remotely via cable, maximum distance of 10 km across. It allows the mirroring configuration, this can improve the system fault tolerance. The size of the server system will be more flexible. Each SCSI channel can connect up to 15 devices, and optical fiber arbitration loop devices can connect up to 126.

    Optical cluster system consists of:

    HP optical cluster hardware including two HP servers (the need to support fiber card, there are LC2000, LH3000, LH4, LH6000, LT6000, LXr8000, LXR8500) and fiber adapter card, you can use RS/12FC optical disk array cabinet are subject to a pair or two pairs of cards for the heartbeat detection and with the client connection. Also needed in the configuration process matching card to the optical fiber optical fiber cable storage device.

    Hardware configuration:

  • HPL system, two web servers
  • Server operating system hard drive 2
  • HP Fibre Array Storage Cabinets (RS/12FC) a
  • Optical disk cabinet dedicated fiber optic cable 2
  • Optical data storage hard disk cabinet, as the user needs to determine
  • HP Fiber Adapter 2
  • NIC 2 web services

    6, the cluster software configuration

    Cluster software on NT platform

    Microsoft, MSCS, there are many third-party professional software developed cluster software, such as Howe's DATAWARE, VINCA company STANDBY SERVER, NSI's DOUBLE-TAKE.

    MS WolfPack features

    MS WolfPack is another name for MS Cluster server, Microsoft Cluster technology developed for two-machine software. It is integrated in the NT SERVER, the support from the machine consisting of two-machine system, to provide a high availability and easy management of applications.

    Main features:

  • Automatically detect and repair errors in server or application
  • Can realize the switching application server
  • Via TCP / IP connections of various clients, such as MS-DOS, WINDOWS 3.X/9X/NT, Apple Macintosh, UNIX, etc.
  • Production host automatically without human intervention to restore the data and take over the task of


  • Automatically review the work of the server and application status
  • Can create high availability applications, file sharing, print requests, etc.
  • Flexibility to set application and data recovery strategy
  • Simple operation can be carried out off-line applications, re-line, the server room migration.

    Currently, WINDOWS 2000 Advanced Server with WINDOWS 2000 DataCenter Server is integrated with more advanced clustering technology.

    Other network operating system platform, there are many cluster software, such as:

    Novell platform based on the cluster software has Novell HA Server, Novell SFT III
    Sco UNIX platform based on the cluster software has a Linux-based cluster software platform Sentinel cluster software has TurboCluster

    7, cluster technology trends

    Clusters with the server hardware and network operating system's development will be the availability, reliability, system redundancy and so gradually increase. The future of the cluster can rely on cluster file system to realize the system all the files, equipment and network resources of the global access, and generate a complete system image. Thus, regardless of the application in which the cluster server, cluster file system allows any user (remote or local) can access this software. Any application can access any file in this cluster. Even in the application from one node to another node in the case, without any changes, the application can access files on the system.

    Today, use of server clustering technology, through careful planning and network maintenance, system failure probability is very small. Therefore, the stability of server clustering technology must be used.