Server Component explanation and analysis

To obtain a server that fully meets your business needs, you must start with the components that provide performance, capacity, and stability for the servers currently under heavy load.

Motherboard

The server motherboard controls all other parts of the server. To provide the performance and scalability required by the enterprise, one or more Pentium (r) high performance or Pentium II processors and Their supported chipset, extended memory and slots should generally be included.

The processor provides computing performance for operating the operating system and any application to control the interaction between various server components. The chipset controls memory, serial and parallel ports, as well as disks and PCI/ISA bus I/O, and processes interrupt requests and DMA requests. A powerful processor or a server with 2-4 processors can provide the performance required by application servers and large networks. It can also increase the flexibility of the system by adding more processor capabilities as needed.

In special servers, the main board may be a SCSI interface highly integrated with shape, support for Disk control, and networking functions. High Integration eliminates the need to control slots on the motherboard. The dedicated server also provides more open slots than the desktop board. Therefore, an ideal server should not only meet your current needs, but also save your long-term costs. This is achieved by providing you with a larger system space that can be scaled as required, rather than forcing you to purchase additional servers ahead of schedule.

If you want to use a server as an intensive disk I/O (such as an Internet server), the peripheral interface bus is a key component. Enhanced IDE provides an economical way to add hard drives, optical drives, and tape drives. The Ultra-widescⅱ/O bus provides higher performance (40 MB/sec), with each port supporting up to 15 peripherals and more than 2800 peripherals. Theoretically, the server board can provide both SCSI (Ultra Wide) and EIDE ports and support circuits, so that you can use the most suitable peripheral interface bus in various situations.

The motherboard design should also include devices that collect information about functions such as temperature, voltage changes, and fan failures. These sensors are usually controlled by a hardware ASIC on the motherboard. As we have seen in the manageability discussion, such devices are crucial for managing servers and maintaining the normal running status of the entire network.

Chassis

The chassis is equipped with a motherboard, power supply, fan, and storage devices. The main problems of the chassis include:

Temperature. How can I cool the internal components of the chassis? The fan (usually from the power supply) should be directly blown or blown through the processor .)

Scalability. Select a chassis with sufficient space for network expansion. Make sure you have enough space to add more hard disks and other peripherals. Check the inside of the chassis to make sure there are no sharp edges. In future upgrades, this is important for technicians to ensure security and prevent accidental disconnection of cables. The chassis should also allow maintenance personnel to easily access internal components for future upgrades or troubleshooting.

Security: Because your server will save important files and more expensive items on the network, you need to consider security issues. The server chassis should be locked to prevent unauthorized dismounting.

Authentication: select a chassis and power supply that meets the requirements of your environment. Typical certifications include FCC, UL, and CE.

Memory

Since data passing through the server memory is very important at any time, it is wise to choose a memory that is verified to run in coordination with the motherboard and tested in extreme conditions. In addition, when mixing different memory, we must follow the advice of the motherboard manufacturer.

The Error Correction Code (ECC) circuit is very important to the server and should be part of the dedicated server. The data in the ECC memory check can be corrected without interrupting the system. Multiple-bit faults are reported, which usually causes system interruption. This is an expensive solution because this feature can be implemented in a memory model (SIMM or DIMM. Another more cost-effective solution is to implement ECC on the memory controller of the motherboard, so that the memory model with lower prices can be applied.

How much memory do you need? For file/print servers, 32 MB memory is enough for each processor. For applications that run a large order input database or other computing-intensive programs, you need to purchase the maximum memory allowed by the budget to improve performance and response time.

Hard Disk and Controller

The hard disk controller can be integrated into the motherboard, which saves both money and space. If the controller is not integrated into the motherboard, you need a controller plug-in card to control the hard disk. Select a controller that provides bus control to increase the data transmission rate.

As hard disks are the most common component that causes server faults, you should choose high-quality hard disks produced by reputable companies with good warranty. Compare MTBF (average fault interval) data to select the hard disk.

To minimize the chances of damage to key data, some servers use the copy hard disk image data. When data is copied to the primary disk array, a data image is also copied to the secondary disk array. Images are highly fault tolerant, but they are expensive because they occupy double disk space to store the same data. You can use a hard disk controller or an operating system image such as Novell sNetWare * or MicrosoftWindowsNT * Server. XmMG
RAID (Redundant Array of Independent Disks) can provide a more cost-effective way to establish fault-tolerant data protection. RAID has three or more hard disks to form a logical drive or array. In RAID 5, a large data block is divided into smaller data blocks and written to different disks. Add the parity bit for the data. When these small data blocks are combined again, you can use the parity bit to check whether the data is correct. If the data is damaged, the data is rebuilt. Like images, RAID can be implemented through hard disk controllers or some operating systems. (For more information about RAID types, visit www.acna..com/raid.html .)

Availability, that is, fast read/write, so that the disk can be used for the next request capability. This is also important when selecting a hard disk. Select a hard disk with a short tracing time. (Seek time is a metric used to measure the time spent reading or writing data to a hard disk .) Adding RAID 5 to the system can also greatly improve the disk read/write speed. The user does not write data to one hard disk at the same time, but to all the hard disks on the server at the same time.

Nic

When the server needs multiple NICs, the ideal solution is to integrate the functions into the motherboard. This improves reliability and integration, and reduces the cost of building servers.

If you have less than 12 network clients, it is sufficient for the server to configure a nic. In a relatively large network, more clients can generate more network traffic. To maintain a high response speed, you can add additional NICs to the server, add additional hubs to the network, and not divide the client into several smaller groups, in this way, each group can connect their respective NICs to the server. U* GR-Z
The network adapter and the network hub jointly control the communication between office computers. Therefore, it is very important to carefully study the options that can be used for these components. You can contact the producer to obtain its specifications, evaluate performance indicators, and make the right choice based on the data, not just the price.

Disk backup

Because the data stored on the server is very important, you need to back up the data to avoid the unfortunate occurrence of data loss. When considering how many large-scale tape backups are to be implemented, remember to ensure the future expansion of hard disk capacity. A m tape backup device seems to be enough to back up the current 2 GB hard disk (if the hard disk space is full, three tapes are required), if you add more hard disks, you will need more tapes. Backing up on a low-capacity tape drive will increase the server backup time.

Therefore, it is meaningful to deploy a tape drive device with a larger capacity than the total capacity of the hard disk to be backed up. In this way, if you back up data at night, you only need to insert a tape, start the backup, and then you can go home. This can also minimize the number of tapes that will increase with the increase of hard disks in the future.

Optical Drive

Because most of the NOS and many applications are provided with CDs, the optical drive is almost indispensable. Other computers on the network can also share the optical drive on the server. If you want to use the server optical drive in this way, we recommend that you use at least 6X or 8X Optical Drive. The SCSI optical drive is easier to set when the EIDE/IDE or SCSI drive is running.

Video Display

Video display capabilities are usually built on the main board to reduce system costs. If not, a cheaper video card is enough, because most server videos are only used to start the system and manage certain system components.

Network Operating System (NOS)

Network Operating System (NOS) is a very important part of the server. The topic of discussing different NOS systems is beyond the scope of this article. The two most common network operating systems today are Novell's NetWare * and Microsoft's WinsowsNT *. Other NOS systems also include IBM's WarpServer *, OS/2, and various versions of UNIX *. For servers with multiple processors, make sure that the selected NOS supports multiple processing.