Broadly speaking, a server refers to a computer system that can provide certain services to other machines in the network (if a PC provides ftp services, it can also be called a server).

In a narrow sense, servers refer to some high-performance computers that can provide services to the outside world through the network. Compared with ordinary PC, it requires higher stability, security and performance, so it is different from ordinary PC in CPU, chipset, memory, disk system, network and other hardware.

Server parsing

As a node of the network, the server stores and processes 80% of the data and information on the network, so it is also called the soul of the network. Figuratively speaking, the server is like a post office switch, while fixed or mobile network terminals such as microcomputers, notebooks, PDA's and mobile phones are like telephones scattered at home, various offices and public places. In our daily life and work, we communicate with the outside world by telephone, and we must go through the switch to reach the target phone; Similarly, network terminal devices, such as microcomputers in families and enterprises, have to go through the server to surf the Internet, get information, communicate with the outside world and entertain, so it can be said that the server is "organizing" and "leading" these devices.

It is a high-performance computer that provides various services for client computers on the network. Under the control of the network operating system, it provides hard disks, tapes, printers, modems and various special communication devices connected with them to client sites on the network, and also provides centralized computing, information dissemination and data management services for network users. Its high performance is mainly reflected in high-speed computing ability, long-term reliable operation and strong external data throughput.

The structure of server is basically similar to that of microcomputer, including processor, hard disk, memory, system bus and so on. They are specially formulated for specific network applications, so there are great differences between servers and microcomputers in processing capacity, stability, reliability, security, scalability and manageability. Especially with the progress of information technology, the role of the network is becoming more and more obvious, and the requirements for the data processing ability and security of your own information system are getting higher and higher. If you are stolen by hackers in the process of e-commerce, you lose key business data; If you can't access the ATM normally, you should consider the behind-the-scenes commander of these equipment systems-the server, instead of complaining about the objective conditions such as the quality of the staff.

Server classification

At present, according to the architecture, servers are mainly divided into two categories:

Non-x86 servers include mainframes, minicomputers and UNIX servers, which use RISC (Reduced Instruction Set) or EPIC processors and mainly use special operating systems such as UNIX. Reduced instruction set processors mainly include IBM's POWER and PowerPC processors, and SPARC and EPIC processors jointly developed by SUN and Fujitsu are mainly Itanium processors jointly developed by HP and Intel. This kind of server is expensive and closed, but it has good stability and strong performance, and is mainly used in the core systems of large enterprises such as finance and telecommunications.

X86 server: Also known as CISC (Complex Instruction Set) architecture server, which is commonly referred to as PC server, it is a server based on PC architecture, using Intel or other processor chips compatible with x86 instruction set and Windows operating system, such as IBM's System x series servers and HP's Proliant series servers. Low price, good compatibility, poor stability and insecurity, mainly used in small and medium-sized enterprises and non-critical business.

From the current network development, the PC server based on x86 architecture has been widely used because of its "small, intelligent and stable" characteristics.

Theoretically, the server is a high-performance computer in the network environment. It listens to service requests submitted by other computers (clients) on the network and provides corresponding services. To this end, the server must have the ability to undertake services and ensure service quality.

However, this explanation is still profound and vague. In fact, servers and personal computers have similar functions, both of which are tools to help human beings process information, but their positioning is different. Personal computer (PC for short) is designed to meet the multi-functional needs of individuals, while the server is designed to meet the needs of many users to process data on it at the same time. How else can many people use the same server at the same time? This can only help to achieve the purpose of * * * with the Internet.

Let's take a look at the function of the server. The server can be used to build web services (the data of web pages we usually see online are stored on the server for people to access), mail services (all the mails we send need to be processed, sent and received by the server), and files can be shared. Print * * * access services, database services, etc. Moreover, these applications have a * * * feature. They are not facing one person, but many people, and they are processing a lot of data at the same time. So the server and the network are inseparable. It can be said that without the network, there would be no server; Servers are born to provide services, and only in the network environment can they exist. The personal computer can completely complete the data processing task of the owner in the case of a single machine.

Server hardware

In fact, the hardware structure of the server system has many similarities with the computers we usually contact. The main hardware structure still includes the following main parts: CPU, memory, chipset, I/O bus, I/O equipment, power supply, chassis and related software. This has also become our main concern when purchasing servers.

The whole server system is like a person, the processor is the brain of the server, all kinds of buses are like nerves distributed in the muscles of the whole body, the chipset is like the spinal cord, and I/O devices are like the hands, eyes, ears and mouth dominated by the nervous system. The power supply system is like a blood circulation system, which transfers energy to all parts of the body.

For a server, the performance design goal of the server is how to balance the performance of each part and make the performance of the whole system optimal. If a server can handle 1000 service requests per second, but the network card can only accept 200 requests, while the hard disk can only carry 150 requests, and the load capacity of various buses can only carry 100 requests, then the processing capacity of this server can only be 100 requests per second, which wastes 80% of the processor.

Therefore, the ultimate goal of designing a good server is to balance the performance of all aspects, so that all parts can cooperate properly and give full play to their respective abilities. We can measure whether the server has achieved the design purpose from these aspects; R: reliability-reliability; A: Availability-availability; S: scalability-scalability; U: usability-ease of use; M: Manageability-Manageability, that is, the RASUM metric of the server.

Because the server provides services in the network, the quality of this service is very important for the network computing environment that undertakes a variety of applications, and the computer hardware that undertakes this service must be able to guarantee the quality of service. This service must first have a certain ability to respond to a reasonable number of server requests in a unit time. At the same time, this service should respond to a single service request as quickly as possible, and this service should always exist within the required time range.

If a WEB server can only handle 1 homepage requests in 1 minute, then other requests except 1 must be queued, and this request must be processed within 3 minutes. At the same time, this WEB server 1 hour ago can be accessed, and it will not be connected after one hour. This kind of WEB server is unimaginable in the current Internet computing environment.

At present, the WEB server must be able to handle thousands of visits at the same time, and the response time of each visit should be short, so the WEB server can't stop, otherwise the WEB server will cause the loss of visiting users.

In order to meet the above requirements, the server hardware must have the following characteristics: performance, so that the server can handle a considerable number of server requests in a unit time, and ensure the response time of each service; Reliability, so that the server can not stop; Scalability enables the server to continuously improve its performance with the increase of the number of users. So we say that we can't use ordinary PC as the server, because PC can't meet the above requirements. In this way, we add another point to the concept of server, that is, the server must have the ability to undertake services and ensure service quality. This is also the main aspect that distinguishes low-cost servers from PCs.

In information systems, servers are mainly used for database and Web services, while PCs are mainly used for desktop computing and network terminals. The fundamental difference in design points determines that the server has more reliable continuous operation ability, stronger storage ability and network communication ability, faster fault recovery function and broader expansion space than PC. At the same time, applications that are sensitive to data also need the server to provide data backup function. On the other hand, PC pays more attention to the usability of man-machine interface, image and 3D processing ability and other multimedia performance.

Server memory

Among the hardware conditions that restrict server performance, memory can be said to be the most important! Its performance and quality is also an important aspect of evaluating server products. However, for server memory, I believe most people still lack understanding because they have little contact. This article mainly answers two questions for readers: What is server memory? What is the essential difference between it and desktop memory?

Explain the importance of server memory.

The server runs the key business of the enterprise, and the downtime caused by a memory error will make the data permanently lost. As an electronic device, memory is prone to various errors.

Therefore, faced with the pressure of enterprise facts and their own shortcomings, manufacturers have already actively introduced their own unique server memory technologies, such as HP's "online backup memory" and hot-swappable mirror memory; IBM's ChipKill memory technology and hot replacement and hot addition memory technology. With the expansion of enterprise information system, the density and capacity of memory will develop accordingly.

Discussion on the performance of server memory

* Server memory is also memory, which is not essentially different from the common memory we usually see in computer city in appearance and structure. It mainly introduces some new technologies into memory, and we can't just draw any conclusions from the appearance. So maybe you're worried. What if someone else sells you the memory of an ordinary PC as server memory? This is generally reassuring, and the possibility is almost zero. Because the memory of ordinary PC is generally not available on the server, it also shows that the memory of the server can not be replaced by the memory of ordinary PC for the sake of greed and cheapness.

* Now the commonly used server memory mainly includes SDRAM and DDR, and there is also a RAMBUS memory, which is a new generation of storage products with high performance and inter-chip interface technology. DDR2, which has just emerged, has gradually extended to server memory. Hyundai Hynix began mass production of DDR2 memory for servers and workstations in June last year.

* From a technical point of view, it is different from ordinary memory because of ECC. This is the abbreviation of error checking and correcting. It is widely used in computer instructions in various fields. ECC is similar to parity. However, in the case that parity can only detect errors, ECC can actually correct most errors. After memory error correction, the computer's operating instructions can continue to be executed. This virtually ensures the stability and reliability of the server system. But ECC technology can only correct unit memory errors. When a multi-bit error occurs, the ECC memory will generate a non-maskable interrupt (NMI) and the system will automatically stop running.

Product understanding

For general memory, users are very concerned about its parameters, such as bandwidth, memory bus speed, waiting period, CAS delay time and so on. But for servers, we often consider the manufacturing process of memory. Server memory generally uses 8 layers of PCB, and the perfect power layer and wiring layer fully reflect the stability gap. As well as the packaging technology of the memory, it can not only bring ideal expansion of volume and capacity to the memory, but also solve the problems of heat dissipation, reliability and density. Manufacturer's products that do well in these aspects, such as:

Kingston server memory. Kingston, as the largest and most professional memory manufacturer in the world at present, with its long-term accumulated experience, makes it equally outstanding in the manufacture of high-end server memory, and has incomparable production technical advantages compared with other memory manufacturers. It provides efficient and stable products for your server with high quality and rigorous testing. In order to ensure the quality of each server memory product, Kingston Company implements a comprehensive quality control process for all memory products, and strictly tests each storage unit of each chip on each server memory product, thus making the reliability of Kingston server memory close to 100%. The following figure is a mainstream product of Kingston, 512mb (PC-133/SDRAM/e), and a single Samsung100gbpc2100ECC DDR. This note is mainly aimed at the entry-level server market. It uses Samsung's own memory chip, model K4H5 10638E-TCB0. Chip capacity up to 64M, 4 Bank architecture design, SSTL2 interface, 66-pin TSOP2 package. The default working voltage is 2.5V, the default working frequency is DDR200 when CL=2, and DDR266 when CL=2.5 and above. The physical structure of the chip is different from the DDR memory chip we usually use, which consists of two layers of chips.

Central processing unit

Server CPU, as its name implies, is the CPU(Center Process Unit Central Processing Unit) used on the server. Readers who have come into contact with local area network must know that the server is an important device in the network and needs to be accessed by thousands of users. Therefore, there are strict requirements for the server, such as fast throughput, super stability and long-term operation of big data. Therefore, CPU is compared to the "brain" of a computer, and it is also the primary indicator to measure the performance of a server. This paper briefly analyzes the products of two CPU manufacturers, aiming to let readers know something, which is different from ordinary CPU!

Let's first look at some features of the server CPU. At present, the CPU of the server is still divided according to the instruction system of CPU, which is usually divided into CISC CPU and RISC CPU. Later, a 64-bit VLIM (Very Long Instruction Word) instruction system CPU appeared.

CISC CPU

CISC is the abbreviation of "Complex Instruction Set Computer" in English and "complex instruction set" in Chinese. Refers to x86 series CPU produced by Intel and its compatible CPU (CPU produced by AMD, VIA and other manufacturers). It is based on PC (Personal Computer) architecture. This kind of CPU is generally 32-bit structure, so we also call it IA-32 CPU. (IA: Intel architecture, Intel architecture). At present, CISC CPU mainly includes Intel's server CPU and AMD's server CPU.

RISC CPU

RISC is the abbreviation of English "Reduced Instruction Set Computing" and Chinese "Reduced Instruction Set". It is developed on the basis of CISC instruction system. Compared with CISC CPU, RISC CPU not only simplifies the instruction system, but also adopts a structure called superscalar and super pipeline. At the same frequency, the performance of RISC architecture CPU is much higher than that of CISC architecture CPU, which is determined by the technical characteristics of CPU. RISC CPU is incompatible with Intel and AMD CPU in software and hardware.

Then let's take a look at the server products of two common CPU manufacturers in the market:

Intel products. Pentium 4 and Celeron are used for personal computers, while Xeon, Xeon MP and An Teng are used for workstations and servers. In addition, although there are some subtle differences in the maximum working frequency, supported FSB and cache capacity of each variety, the internal design is basically the same, which ensures the compatibility of the software. The biggest difference between Pentium 4 (or Celeron) and Xeon is that it supports "multi-processor system", and multiple CPUs are installed in one machine. Pentium 4 can only install one physical CPU in the whole system, while Xeon can integrate two, and XeonMP can even integrate more than four. Itanium processors should be specifically mentioned here:

This kind of processor should be said to be one of the processors that most people don't know much about. Because it is specially designed for demanding enterprises and technical applications, aiming at the high-end enterprise market, and compared with other Intel series processors, it is expensive, even the cheapest model is still worth more than $ 1000! Itanium processor is based on IA-64(Intel Architecture 64), and 64-bit is just a technical feature of Itanium processor. The latest An Teng 2 processor has a system bus bandwidth of 6.4GB/ s, an integrated L3 cache of 6MB, and a main frequency of 1.5GHz.

AMD products. AMD has been eager to try in the server field since 200 1. In June, it launched Althlon MP supporting dual processors and AMD-760 MP chipset supporting DDR ECC SDRAM and AGP 4X. Chipset includes AMD-762 system controller (North Bridge) and AMD-766 peripheral bus controller (South Bridge), but AMD-762 only supports 33MHz 64-bit PCI. AMD Athlon MP processor is compatible with stable and reliable AMD Socket A structure and can support DDR memory. This processor is manufactured by AMD's 0. 13 micron copper wire technology and produced by AMD's Fab 30 chip factory in Dresden, Germany.

AMD Athlon MP processor is the latest model of AMD Athlon series processors, which can ensure the unprecedented high performance of multiprocessor systems. This processor is the first seventh-generation x86 processor with this capability in the world, which can support servers and workstations on high-performance multiprocessor platforms. M MP processor is an essential component for configuring single-processor (1 way) and dual-processor (2-way) server and workstation platforms, especially suitable for commercial and enterprise systems. This processor has a unique design, which is most suitable for applications that execute multi-threads and are responsible for important tasks.

Rack server

Rack servers look like switches rather than computers, and the specifications are1u (1u =1.75 inch = 4.45cm), 2U and 4U. Rack servers are installed in standard 19-inch cabinets. Most of these structures are functional servers.

For information service enterprises (such as ISP/ICP/ISV/IDC), physical parameters such as server size, power consumption and calorific value should be considered first when selecting servers, because information service enterprises usually use large dedicated computer rooms to uniformly deploy and manage a large number of server resources, and the computer rooms are usually equipped with strict security measures, good heat dissipation systems and multiple standby power supply systems, and their computer rooms are quite expensive. How to deploy more servers in a limited space is directly related to the service cost of enterprises. Usually choose a rack server with a mechanical size of 19 inch. Rack servers also have various specifications, such as 1U (height 4.45cm), 2U, 4U, 6U, 8U and so on. General 1U rack server saves the most space, but its performance and scalability are poor, so it is suitable for some application fields with relatively fixed business. Products above 4U have high performance and good scalability, and generally support more than 4 high-performance processors and a large number of standard hot-swappable components. Management is also very convenient. Manufacturers usually provide people with corresponding management and monitoring tools, which are suitable for key applications with a large number of visits, but they are bulky and have low space utilization.

Cabinet server

In some high-end enterprise servers, due to the complex internal structure, there are many internal devices, and some have many different equipment units or several servers in a cabinet. This server is a cabinet server.

For important enterprises such as securities, banks, posts and telecommunications, systems with complete self-repair ability should be adopted, and redundant measures should be taken for key components. For the servers used in critical business, you can also use a dual-machine hot standby high-availability system or a high-performance computer, which can well ensure the availability of the system.

Server procurement

In the eyes of ordinary computer users, servers always appear mysterious. With the popularization of network environment, servers are used more and more, and ordinary users have more and more opportunities to contact servers. This paper aims to uncover the mystery of the server.

The concept of server

The English name of a Server is "server", which refers to a special computer that provides some services for clients under the network environment. The server is equipped with network operating system (such as Windows 2000 Server, Linux, Unix, etc.). ) and various server application system software (such as Web services and e-mail services). The "client" here refers to the computer installed with operating systems used by ordinary users such as DOS and Windows 9x.

The processing speed and system reliability of the server are much higher than that of the ordinary PC, because the server works continuously in the network. Ordinary PC crashes, restarting is a big deal, and data loss is limited to a single computer.