Managed by the system administrator structured computer environment and only one or a few isolated computer components computer environment The main difference is the service. This only a few isolated computer environment for families and those typical small non-technical office , and typical computer environment is structured by the technical personnel to operate a large number of computers to facilitate communication by sharing , optimize resources and other services to interconnected together. When a home computer through the Internet or through an ISP to connect to the Internet , he is using an ISP or services provided by others before entering the network .
由系統(tǒng)一管理理員管理的結(jié)構(gòu)化計算機背景和只有一臺或幾臺孤立計算機組成的計算機背景的主要差別是服務(wù)。這種只有幾臺孤立計算機的典型背景是家子和那一些細(xì)小的非技術(shù)性的工作室，而典型的結(jié)構(gòu)化計算機背景則是由技術(shù)擔(dān)任職務(wù)的人操作準(zhǔn)量多的計算機，經(jīng)過共享便捷的通信、優(yōu)化的資源等服務(wù)來相互連接在一塊兒。

Provide a service more than just a simple accumulation of the hardware and software together, it includes service reliability , service standardization , as well as service monitoring, maintenance , technical support and so on. Only in these areas are in line with the requirements of service is the real service .
供給一個服務(wù)絕不止只是簡單的把硬件和軟件累加在一塊兒，它涵蓋了服務(wù)的靠得住性、服務(wù)的標(biāo)準(zhǔn)化、以及對服務(wù)的監(jiān)控、保護(hù)、技術(shù)支持等。只有在這幾個方面都合乎要求的服務(wù)才是真正的服務(wù)。

A service of the basic problems .
Create a stable and reliable service is a system administrator's important work. In doing so , the system administrator must consider a number of basic elements , the most important thing in all stages of design and development must take into account the needs of users . And users to communicate , to find the user 's requirements and expectations of the service , and then the other requirements such as management requirements, etc. Make a list , so the list can only let people see the system administrator team .

Services should be based on server-class machine and the machine should be placed in a suitable environment , as the server machine should have an appropriate reliability. Services and services that depend on the machine should be monitored in the event of failure or malfunction on the alert list of records .

As part of the service should rely on those machines and software built on the same or higher standards and software on the host , a service reliability and service it depends on the weakest link in the chain is quite reliable . A service should not be a reason to rely on those who are not part of the service host. Once the service built and completed a test , it is necessary gradually to the user's point of view to further testing and debugging.

1.1 user requirements.
To establish a new service should start from the user's requirements , you create a service user is the root cause . If the service does not meet the established needs of users , and that is a waste of energy.
Collect user requirements should include the following contents : how they want to use these new services need to function in , like what features , how important these services to them , and what they need for these service levels of availability and technical support. If possible , allow users to try out the service trial version . Do not allow users to use those very troublesome or unsuccessful systems and projects. Try to calculate the user group to use this service and how much they need and want what kind of performance , so as to correct calculation .

1.2 operational requirements.
For system administrators, the new service some requirements are not directly visible to the user . Such as the system administrator to take into account the new service management interface , it can interoperate with existing services , as well as whether the new service with core services such as authentication services and directory services such as integrated together.

From the user's desired level of reliability and system administrators required reliability of the system in the future expectations, system administrators can create a list of features that users expect , which includes a cluster , the slave device , the backup server or high availability hardware and operating systems.

1.3 open architecture .
Sometimes vendors use proprietary protocols and other vendors is to reach a definitive license agreement, but will be used in a new version of vendors and other vendors use a compatible version of a noticeable delay between the two vendors used there will be interrupted between the versions , but did not provide the interface between the two products . This situation for those who rely on their interfaces using both products of the people, is a nightmare .

A good solution is to choose based on open standard protocols , so that both sides can choose their own software . This brings the choice of the client application server platform selection process with the separation of the user freedom of choice that best suits their needs , preferences and even the platform software , the system administrator can also independently chosen based on their reliability , scale configurability and manageability required server solutions . System administrators can in some competing server products to choose from , rather than being confined to those suitable for some client server software applications and platforms .
In many cases , if the software vendor supports multiple hardware platforms , the system administrator can even independently selected server hardware and software.

We call the user to select and server selection separation capability. Open protocol provides a fair competition venues, and provoke competition between vendors , which will ultimately benefit us .
Open protocols and file formats is quite stable , does not often change ( even change is upward compatible ) , but also a wide range of support , can give you the largest independent product selectivity and the greatest opportunity to obtain reliable , good compatibility product 2 Other issues to be considered .

In addition to requiring a reliable service can be monitored , easy maintenance support, and we want to meet all the basic requirements and the requirements of users , but also take into account some special things. If possible , it should be allowed to use each service dedicated machine , so as make it easier for services can be supported and maintained , but also can reduce forget some small server machine services. In some large companies, the use of specialized machines are a fundamental principle , while in small companies , due to cost issues , generally meet this requirement .

Another idea is to create a service to allow the service to be fully redundant as the goal. Some important services regardless of how much companies are required to fully redundant . As the size of the company will continue to grow , all you have to let all the services are fully redundant as the goal.

2.1 the use of specialized machines.
Ideally , services should be based on a special machine.
Large sites should have the ability to adjust according to service requirements to this structure , and small sites are difficult to achieve. Each service has a dedicated machine will make services more reliable , when a reliability problem is also easy to debug , failure range is smaller, and easier to upgrade and perform capacity planning .
Small companies to grow up big sites generally have a centralized management of the machine as the core of all the important services , this machine provides naming service , authentication services , print services , mail services and so on. Finally, since the load growth , the machine had to be separated , to extend the service to other server up . Often before this, the system administrator who has got money, you can buy more for management of the machine, but found it too troublesome because there are so many services rely on these machines , they are too difficult to separate . When the service from a machine apart , IP address dependency difficult to deal with , and some services, such as the name of the service IP addresses where the user already remember very strong , there are some IP addresses are safety systems such as routers, firewalls and other uses.
To a central host decomposed into many different hosts is very difficult to build up the longer , the above services , the more difficult it exploded . Name -based services will be helpful, but it must be the entire company using standardized , uniform , consistent names.

2.2 sufficient redundancy.
Full redundancy is a good one or a series of replicated servers, the time in the event of failure to take over the main faulty equipment . Redundant systems should be able to run continuously as a backup server if the primary server fails to automatically connect to the line, or just a small amount of human intervention , we can take over the failed system of services provided .

You choose this type of redundancy is dependent on the service . Some services such as web servers and computing area, you can make yourself a good run on the cloning of a good machine . Other services such as large databases will not work, they require a stronger connection crash recovery system . You are using the software may be used to provide the service will tell you that redundancy is an effective , passive form from the server , only the primary server fails and sends a request , the redundant system will response . Whatever the case , the redundant data synchronization mechanism must ensure and maintain data integrity.

If redundant servers and the primary server synchronization continuous running, redundant servers can be used to share the load of running normally and can improve performance. If you use this method , we must be careful not to load beyond the critical point of unacceptable performance , in order to prevent a server malfunction. Before reaching the critical point for the existing system to add more parallel servers .

Another advantage is redundant easy to upgrade . Rolling upgrade can be performed . Each time a host is disconnected , upgrading, testing and start services. Single host failure will not stop the entire service , although it may affect performance. If you really messed up an upgrade then wait for you to calm down and turn off the power to go fix it .