Linux server cluster system implementation solution detailed analysis

There is a common way to greatly improve the security of the server, this is the cluster.

Cluster cluster technology can be defined as follows: A set of independent servers behave as a single system in the network and managed in a single system mode. This single system provides highly reliable service to customer workstations.

In most modes, all computers in the cluster have a common name, and services running on any system in the cluster can be used by all network clients. Cluster must coordinate the management of errors and failures of separate components and transparently add components to the Cluster.

A Cluster consists of multiple (at least two) servers with shared data storage. When any server runs an application, the application data is stored in the shared data space. The operating system and application files for each server are stored in their respective local storage.

Each node server in the cluster communicates with each other through an internal LAN. When one node server fails, the application running on this server will be automatically taken over on the other node server. When an application service fails, the application service will be restarted or taken over by another server. When any of the above failures occur, the customer will be able to connect to the new application service very quickly.

Second, the hardware configuration of the cluster

Mirror server dual-machine

The mirror server dual-system in the cluster is the simplest and cheapest solution for hardware configuration, usually mirroring The hardware configuration of the service requires two servers. Each server has a separate operating system hard disk and a data storage hard disk. Each server has a network card connected to the client, and a pair of mirror cards or network cards that complete the mirroring function.

The mirror server has many advantages, such as simple configuration, convenient use, and low price. However, since the mirror server needs to use the network mode to mirror data and synchronize the data through the mirroring software, it needs to occupy the CPU and memory resources of the network server. The performance of a mirror server is lower than that of a single server.

There are some mirror server cluster systems that use memory mirroring technology. The advantage of this technology is that all applications and network operating systems are mirrored on both servers. When the host fails, the backup machine can be almost Take over all applications without feeling. Because the memory of the two servers is exactly the same, when the system application has a defect and the system is down, the two servers will be down. This is also the memory mirror card or network card to achieve data synchronization, in the process of reading and writing large data, the two servers will generate data out of synchronization in some states, so the mirror server is suitable for those users with less budget and less demanding cluster system. .

Hardware Configuration Example:

Two server operating system hard disks of network server

Server data storage hard disk depends on user needs

Server image card (Some software can use standard network card) Two

Network service network card two three, dual machine and disk array

Compared with the mirror server two-machine system, the two machines are interconnected with the disk array cabinet There are more disk array cabinets produced by third parties. At present, many companies such as Haowei Company and Jingye Company produce disk array cabinets. Disk array control cards are installed in the disk array cabinets. The array cabinets can be directly installed in the cabinets. The hard disk configuration becomes a logical disk array. The disk array cabinet is connected to the normal SCSI card on the server through a SCSI cable. The system administrator needs to configure the disk array directly on the disk cabinet.

The dual-machine and disk array cabinet interconnect structure does not use memory mirroring technology, so it needs a certain switching time (usually 60?D?D180 seconds), which can have suburban avoidance due to the defect of the application itself. The system is completely down, and since all the data is stored in the central disk array cabinet, when the working machine fails, the backup machine takes over the working machine and reads data from the disk array, so no data is generated. The problem of synchronization, because this solution does not require network image synchronization, the performance of this cluster solution server is much higher than the image server structure.

The disadvantage of the interconnection structure between the dual-machine and the disk array cabinet is that there is a single-point error in the system. The so-called single-point error refers to all the failures of a certain component or an application in the system. The system is down. In this system, the disk array cabinet will cause a single point of error. When there is a logical or physical failure of the disk array cabinet, all the stored data will be lost. Therefore, when selecting this solution, you need to choose a quality and after-sales service. A service with better products.

Hardware Configuration Example:

Two Network Servers

Two Operating System Hard Disks

A Disk Array Cabinet Produced by Third Parties

Two SCSI wires for disk cabinets

Disk storage cabinet data storage hard disk is determined according to user needs

Two network service network cards

In addition, some Manufacturers have better technical solutions, such as HP.

3. HP dual-machine dual-control fault-tolerant system

HP NetServer provides high-quality and high-quality dual-controller fault-tolerant systems Reliable hardware foundation...

The HP dual-machine dual-control fault-tolerant system combines the security and reliability of HP server products with the advantages of Cluster technology to complement each other's strengths.

Hardware Configuration Example:

Two Network Servers for HP L System

Two Operating System Hard Disks

HP Hard Disk Storage Cabinet (SS/6, RS/8, RS/12) One

Two SCSI cluster adapter cables for disk cabinets

Disk cabinet data storage disks are determined according to user requirements

HP cluster dedicated array card two

Network service network card two five, HP Fibre Channel dual-machine dual-control cluster system

Fibre Channel is a connection standard, can be used as an alternative to SCSI Solution, fiber technology has high bandwidth, anti-electromagnetic interference, long transmission distance, high quality, strong expansion capability, etc. Currently, 126 devices can be connected to the FC-AL arbitration loop.

Fiber optic equipment offers a variety of enhanced connectivity technologies that are greatly accessible to users. The server system can be remotely connected via fiber optic cable, which can span a distance of up to 10 kilometers. It allows for mirroring configuration, which improves the system's fault tolerance. The size of the server system will be more flexible. SCSI can connect up to 15 devices per channel, while fiber arbitrated loops can connect up to 126 devices.

Fiber Cluster System Components:

HP Fiber Cluster System Hardware Equipment includes two HP servers (requires support for fiber cards, currently LC2000, LH3000, LH4, LH6000, LT6000, LXr8000, LXR8500) And fiber adapter cards, you can use RS /12FC fiber disk array cabinet, plus one or two pairs of network cards for heartbeat detection and connection with the client. In the configuration process, an additional fiber optic cable for the fiber card to the fiber storage device is required. Hardware configuration:

Two network servers of HPL system

Two operating system hard disks

HP fiber array storage cabinet (RS/12FC) One

Two fiber optic cables for fiber-optic disk cabinets

Data storage disks for fiber-optic disk cabinets are determined according to user needs

Two pieces of HP fiber adapter cards

Network Service Network Cards Two

Four, cluster software configuration

NT platform-based cluster software

Microsoft's MSCS, there are also many third-party professional software company developed cluster software, such as Howard's DATAWARE, VIN CA's STANDBY SERVER, NSI's DOUBLE-TAKE

MS WolfPack Features

MS WolfPack is another name for MS Cluster server, which is developed by Microsoft for Cluster technology. Dual machine software. It is integrated on NT SERVER and supports a two-machine system consisting of two machines, providing a highly available and manageable application environment.

Main features:

Automatic detection and repair of server or application errors

Switching to applications in the server

Available via TCP/IP connection to various clients, such as MS-DOS, WINDOWS 3.X/9X/NT, Apple Macintosh, UNIX production host can automatically recover data and take over tasks without manual intervention

Manageability:

Automated auditing of server and application working status

Highly usability applications, file sharing, print requests, etc.

Flexible setting of application and data recovery strategies

The application can be offline, re-wired, and migrated between servers.

At present, WINDOWS 2000 Advanced Server and WINDOWS 2000 DataCenter Server are integrated with more advanced clustering technology.

There are also many clustering software on other network operating system platforms, for example:

Cluster software based on Novell platform has Novell HA Server, Novell SFT III

Based on sco UNIX platform Cluster software with Sentinel cluster software

Linux-based cluster software with TurboCluster

V. Cluster technology trends

Cluster technology with server hardware systems and network operations The development of the system will gradually improve in terms of availability, high reliability, and system redundancy. Future clusters can rely on the cluster file system to achieve global access to all files, devices, and network resources in the system, and to generate a complete system image. This way, the cluster file system allows any user (remote or local) to access the software, regardless of which server in the cluster the application is on. Any application can access any file in this cluster. Even if the application is moved from one node to another, the application can access the files on the system without any changes.

Today, using server clustering technology, through careful planning and network maintenance, the probability of system corruption is very small. Therefore, enterprise server stability must use clustering technology.