Comparison between server memory and normal memory

Server memory is also memory (RAM). There is no obvious substantial difference between it and ordinary PC (personal computer) memory in appearance and structure. It mainly introduces some new and unique technologies in memory, such as ECC, ChipKill, hot-swap technology, etc., which has extremely high stability and error correction performance.

Main server memory technologies:

(1)ECC

In ordinary memory, a technology called Parity is often used. Parity check codes are widely used in error detection codes. They add a check bit to each character (or byte) of data and can detect all odd (or even) parity errors in a character. However, Parity has a disadvantage. When the computer finds that a byte has an error, it cannot determine which bit the error is in, and thus cannot correct the error. Based on the above situation, a new memory error correction technology has emerged, that is, ECC. ECC itself is not a memory model, nor is it a memory-specific technology. It is a computer instruction widely used in various fields and is an instruction error correction technology. The full name of ECC in English is “Error Checking and Correcting”, and its corresponding Chinese name is “Error Checking and Correcting”. From the name we can see that its main function is to “detect and correct errors”. It is more advanced than parity correction technology mainly in that it can not only detect errors, but also correct them. Only after these errors are corrected can the computer correctly perform the following tasks and ensure the normal operation of the server. The reason why it is not a memory model is that it is not a technology that affects the memory structure and storage speed. It can be applied to different memory types, just like the "parity correction" memory mentioned earlier. It is not a type of memory. The first to apply this technology was EDO memory, and now SD also has applications. ECC memory is mainly widely used starting from SD memory, and the new DDR and RDRAM also have corresponding applications. The current mainstream ECC memory is actually a type of SD memory.

(2) Chipkill

Chipkill technology was developed by IBM to address the shortcomings of ECC technology in current server memory. It is a new ECC memory protection standard. We know that ECC memory can only detect and correct single-bit errors at the same time, but if more than two bits of data are detected to be erroneous at the same time, it is generally powerless. The reason why ECC technology is currently widely used in server memory is that other new memory technologies were not mature before, and secondly, the system speed in current servers is still very high. Generally speaking, it is rare for multiple bit errors to occur at this frequency. It is precisely because of this that ECC technology has been fully recognized and applied, making ECC memory technology the memory standard on almost all servers.

However, as the CPU performance of servers based on Intel processor architecture has increased exponentially, the performance of hard disk drives has only increased by a small multiple during the same period. Therefore, in order to obtain sufficient performance, the server needs a large amount of memory to temporarily store the data that needs to be read on the CPU. Such a large amount of data access results in a single memory chip usually providing 4 (32 bits) or 8 (64 bits) bits of data or more each time it is accessed. Reading so much data at one time greatly increases the possibility of multi-bit data errors, and ECC cannot correct errors of more than two bits. This is likely to cause the loss of all bits of data, and the system will quickly crash. IBM's Chipkill technology uses the memory substructure method to solve this problem. The memory subsystem is designed so that for a single chip, no matter how wide the data is, it affects at most one bit for a given ECC identifier. To give an example, if a 4-bit wide DRAM is used, the parity of each bit in the 4 bits will form a different ECC identification code, and this ECC identification code is saved using a single data bit, that is, saved in different memory space addresses. Therefore, even if the entire memory chip fails, each ECC identification code will have at most one bit of bad data, and this situation can be completely repaired through ECC logic, thereby ensuring the fault tolerance of the memory subsystem and ensuring that the server has strong self-recovery capabilities when a failure occurs. Memory using this memory technology can check and repair 4 erroneous data bits at the same time, and the reliability and stability of the server are more fully guaranteed.

(3)Register

Register is a register or directory register. We can understand its role in memory as the directory of a book. With it, when the memory receives a read or write instruction, it will first retrieve this directory and then perform the read or write operation, which will greatly improve the working efficiency of the server memory. Memory with registers must have buffers, and all register memories currently available also have ECC functions. They are mainly used in mid-to-high-end servers and graphics workstations, such as IBM Netfinity 5000.

Typical types of server memory:

Currently, there are two types of memory commonly used in servers: SDRAM and DDR.