The server is simply a computer, but many people do not understand why the server is different from the computer, which is a point that many people will be puzzled. Today, Micronet will analyze its technical advantages from the memory of the server. Server storage is also memory (RAM). Like general PC (personal computer), its appearance and structure are not significantly different, mainly because it introduces some unique new technologies in memory, such as ECC, ChipKill, hot plug, etc., with high stability and error correction performance.
Main server memory technologies:
(1)ECC
Parity technology is often used in general memory. It is the same kind of Parity check codes that are usually used for error detection codes, It is not possible to determine which bit the error is in and correct it. Based on the above situation, this paper proposes a new memory error correction technology, namely ECC. ECC itself is not a memory model or a special memory technology, but a computer instruction error correction technology widely used in various fields. Its full English name is "ErrorChecking and Correcting", and its matching Chinese name is "Error Checking and Correcting". From this kind of name, we can see that its main function is to "find and correct errors". Compared with parity correction technology, it is more advanced, mainly because it can not only find errors, but also correct them. After these errors are corrected, The computer can correctly execute the underlying tasks and ensure the normal operation of the server. The reason why it is not a type of memory is that it is not a technology that affects the memory structure and storage speed. It can be used in various memory types. Like the "parity correction" memory mentioned earlier, it is not a type of memory. EDO memory was the first one to use this technology, At present, the mainstream memory is ECC memory, which is mainly popularized from SD memory. At the same time, the new DDR and RDRAM also have corresponding uses. At present, ECC mainstream memory is actually a kind of SD memory.
(2)Chipkill
IBM has developed Chipkill technology to better solve the shortage of ECC technology in server memory and serve as a new standard for ECC memory protection. As we all know, ECC memory can only detect and correct single bit errors at the same time, but if it detects errors in two or more upper level data at the same time, there is usually no way. At this stage, ECC technology can be widely applied to server memory. On the one hand, the new memory technology is not mature before this stage, and on the other hand, the system speed of ECC technology on server memory is still very high at this stage. Generally, multiple bit errors are rare, That is to say, ECC memory technology can be fully recognized and used on any server and become a storage standard.
However, in recent years, the CPU characteristics of servers based on Intel processor architecture have increased exponentially, while the characteristics of hard disk drives have only increased several times. This is because in order to obtain sufficient performance, servers need a large amount of memory to store data that needs to be written on the CPU when it comes to time, Such a huge amount of data access makes each single memory chip can only produce data of about 4 (32) bits or 8 (64 bits) bits each time it accesses, so the possibility of multi bit data errors will be greatly increased if multiple data are written at one time, and ECC cannot correct errors of about double bits, which may lead to the loss of all bit data, The system didn't crash so quickly. IBM Chipkill technology uses the memory mode of substructure to deal with such problems. The design principle of the memory subsystem is that a single chip, no matter how wide the data width is, is only related to the ECC identification code, and its maximum impact is 1 bit. For example, if DRAM with a width of 4 bits is used, the parity of each bit in the 4 bits will form a different ECC identification code. This ECC identification code is stored in different data bits, that is, it is stored in different memory space addresses. In order to achieve this goal, all memory chips will fail, and each ECC ID code will have up to one bit of bad data, and this situation basically needs to be repaired through ECC logic, so as to ensure the fault tolerance of the memory system and the strong self recovery ability of the server in case of failure. Using this memory technology requires checking and repairing 4 wrong data bits at the same time, and the reliability and stability of the server are well guaranteed.
(3)Register
Register is a register or directory register. Its role in memory can be parsed into a book directory. When a memory is connected to a read/write instruction, it will first retrieve the directory, and then start the read/write operation. This will greatly improve the productivity of server memory. The storage containing Register must have a Buffer, and the Register storage that can be encountered at this stage also has ECC function, which is specially used for high-end servers and graphic workstations, such as IBMNetfinity5000.
Typical server memory type: At this stage, the memory used by the server mainly includes SDRAM and DDR.
The above points are analyzed by the neutrino network What are the technical advantages of server memory compared with ordinary PCs? I believe you also have a better understanding of servers , if you still don't know anything, please contact us. If you have more questions about virtual hosting, please consult Micronet. Micronet is an IDC service provider focusing on server rental and hosting. With more than 10 years of industry experience, it is safe, stable, reliable and reassuring. It is a leading enterprise in the domestic IDC industry. It helps thousands of enterprises to achieve network informatization, 7 * 24 hours of manual service, after-sales care free, and has a good reputation.
