In the foregoing, we introduced the consideration of so many server leases. Now we must discuss with you how to choose the CPU, motherboard, memory, hard disk and other components of the server.
How to select the server's CPU, motherboard, memory, hard disk
Speaking of server hiring, many readers have expressed embarrassment and said that they have not even seen the real server surface, and they have not even seen it. In fact, this is only a superficial phenomenon in the process of understanding anything. Most of the things that have not been touched for the first time will always feel very mysterious, and then once they are really in touch, the big number will send out "the original" feeling.

There is nothing mysterious about the server. Most of the considerations are similar to those of our own PCs. After all, the overall architecture is similar to that of a PC. It is also composed of a motherboard, CPU, memory, hard disk, and various boards. , external equipment, power supply, chassis and other components. Moreover, the server is still developing from the PC. There is nothing to fear.

1.CPU selection

As with our usual PC purchase, we must first determine the CPU, not the motherboard, because the CPU is the core of the entire server. The performance of the service stick depends to a large extent on the CPU type selected.

To select the CPU, you must first know the current server CPU brand, type and model, at least the series name. In the PC, there are only two brands that we can choose from at present. It is Intel and AMD, but in the area of server processors, the choice of CPU is much larger. In addition to Intel and AMO, there are currently IBM and Sun. The processors of Intel and AMO are based on x86 or IA-64 architecture. The main operating systems supported are Microsoft Windows systems and various brands of Linux systems. The processors of IBM and SUN are all based on RISC architecture. It mainly supports its own UNIX system and some specific brands of Linux systems. On the whole, Intel and AMO's processors have good compatibility and wide support, but generally only meet the needs of middle- and low-end users. Users like IDC, telecommunications, finance, and securities usually choose to support UNIX systems. IBM and Sun processor server systems. In terms of price, of course, servers based on Intel and AMO processors are cheaper.

The determination of the processor brand is only the first step, and the type of processor is also selected later. Among the four major server processor brands, there are single-processors that only support one processor, and dual-processors (DP) that only support two processors. Of course, there are multiple processors that support multiple processors ( MP) too. The specific choice is to fully consider the current and future application requirements for at least 3 years. If the company is relatively small and business development is not fast, consider choosing a processor that supports a single channel, but this is not what the author supports. The author strongly recommends that at least dual-circuit DP processors be supported so that the previous investment can be protected to the maximum. However, it should be noted that it is necessary to know in advance whether the selected DP processor must be installed with two processors (some are possible, some are not) in order to work properly, so as to avoid regret after purchase.

If you choose to support multiple processors, you must understand whether the architecture adopted is SMTP symmetric architecture or asymmetric MMP architecture to purchase a corresponding number of processors based on actual needs.

After the processor brand and type are determined, it is necessary to select a specific model. At present, in the server processors of Intel and AMD, the main competition is the dual-core, quad-core processors, plus the two competitions. The price of dual-core processors is basically the same as that of single-core processors. When selecting the processors of these two companies, it is better to choose the product number of dual-core products. For medium-sized companies and above, or for relatively quickly developing small businesses, it is recommended to choose quad-core or even multi-core processors that will soon be available. In contrast, processors with multi-core technology are more affordable than buying multiple processors. If you choose a quad-core processor and you have four processors installed at the same time, the server can allow 16 simultaneous processes, which is equivalent to 16 single-core processors.

In addition, dual-core, quad-core processor products, there are different series and models, the specific choice is based on the actual needs and economic affordability to be set, do not necessarily have to choose the latest, is the best application.

IBM and Sun's processors are currently dual-core, quad-core and eight-core processors, primarily for supporting their own UNIX operating systems.

2. The choice of motherboard

The next thing to do after selecting the processor is the choice of the server board. The main board is the most direct embodiment of the new server technology, new features, new features, and overall performance. The mainboard of the server is more important than the main board of an ordinary PC because the server main meal not only determines the latest technology adopted by the server, but also determines the overall performance of the service server. Of course, the choice of the server board is first determined by the processor's choice of the front panel. However, unlike PCs, we do not choose the motherboards ourselves in the market, but we purchase them in the form of complete machines. At this time, the performance of the server motherboard is achieved through the introduction of the corresponding model server. It is usually recommended to use the raw pull chip developed by the corresponding brand processor manufacturer. Thin motherboards provide better compatibility and can maximize the use of new processors and new features.

Our main consideration in selecting the mainboard is the standard configuration and expansion performance of the server. Such as the number of processors supported, the type of memory and memory supported, the number of various disk interfaces, various types and number of expansion slots (such as PCI, PCl-X, PCI-E, or LnfiniBand I/O, etc.) Integrated RAID controller card, RAID controller card type, integrated question card type and number, and other internal/external interface configurations.

The above configuration depends on the selected server level. The entry-level server configuration is the lowest, the processor generally only supports 2 channels, the memory is usually only 4GB, and the disk shelf usually only provides 4 I/I. The number of O expansion slots is usually within 6 (usually 32-bit and 64-bit PCI bus), and does not support external expansion. Applicable to small businesses around 50 users.

The departmental server is slightly higher than the workgroup server. The processor supports 4 to 8 channels, the maximum memory is about 12GB, the disk shelf is about 8 and the number of I/O expansion slots is about 8 (usually provided New busses such as PCI-X, PCI-E), and usually external extensions are also supported. A medium-sized enterprise that is applicable to more than 200 users and has a rapid business development.

The enterprise-class server configuration is the highest, usually supporting more than 8 processors. The maximum memory is about 16GB. The disk shelf can provide more than 10 channels through the expansion of the cabinet. Various I/O expansion slots (usually provided like PCI-X, New buses such as PCI-E and lnfiniBand can also be extended by peripheral expansion. Applicable to corporate users over 500 users.

3. The choice of memory

The importance of memory is second only to the CPU, and even more so in the server, because many of the server's failures are caused by memory. In terms of server memory selection, not only current mainstream memory technologies such as DDR2, DDR3, and supported memory capacity should be considered, but also other features it has, such as support for internal memory modules (such as common DIMMs), There are now the latest FB-DIMMs, error correction methods (at least to support the standard ECC), whether to support internal mirroring, memory protection, memory arrays and other technologies.

4. The choice of hard disk

Because the server has to face multiple users' concurrent file access, the read and write performance of the hard disk used by the server is required to be very high. This aspect is ensured by increasing the rotational speed of the hard disk and the bandwidth of the hard disk interface. On the other hand, it is also reinforced by RAID.

With regard to hard disk interfaces, server hard disks still use parallel SCSI. However, the new SATA and SAS interfaces have also begun to be adopted in some high-end servers because these two new interface technologies have matured after years of development. Hard disk speed is usually above 1000OOPMPM.

In terms of RAID, the SC button, RAID, SATA-RAID, and SAS-RAID can also be used in the server area, which is consistent with the three main application hard disk interfaces. In contrast, SCSI-RAID and SAS-RAID have higher performance, and the SAS interface technology is relatively new. Currently, there are fewer server supports and fewer such array card options.

For most small and medium-sized enterprises, they choose standard SCSI interface hard disks and lower-level arrays such as RAIDO, RAID1, and RAID5. Small-business users can also choose hard disks and RAIDO and RAID1 arrays with SATA interfaces: For performance requirements High- medium and large enterprise users can choose the SAS interface hard disk with the highest performance currently and RAID 5 or higher level array.