Subtitles of the Movie

Let's pretend that you have a pile of boxes that you want me to put into a closet. And I am going to use these boxes later on, but I want to make sure that I use only boxes that I put into the closet. Right now the closet is empty. Some of the boxes that you have contain a big red "X" or have on the side of the box a big red "X"; some of them don't. If you hand me a box and it has a big red "X" on it, then I am going to go ahead and put it in the closet. But if you hand me a box that does not have a big red "X" on it, then I am going to take out my red marker and I am going to put a big red "X" on it, and then put it in the closet. If I open the closet door later on and see boxes that do not have red X's on them, I know I didn't put those boxes in there. In the same way, error checking by parity works. In other words, a byte is composed of eight bits. When bytes are put into memory, the number of 1's in the byte are counted. All of the bits either have to be a one or a zero, and there are eight in each byte. If we count the number of 1's then we can setup a standard. For example, if we use even parity, then we are saying that we are always going to have an even number of 1's in our bytes. So if we look at a byte and say it has five 1's and three 0's, and we are using an even parity memory, then memory would go ahead and add another 1 to it, called a parity bit. This makes 9 bits in that byte, but it also makes sure that that byte has an even number of 1's. By the same token, if we are using odd parity and a byte only had four 1's in it, then memory would go ahead and add another bit called a parity bit, to make five 1's so that that byte would contain an odd number of 1's. The reason this is important is that memory degrades over time, actually over milliseconds, and it constantly has to be refreshed. And it's possible to lose a bit. But most memory errors don't go any further than just one bit - so parity checking works in a lot of cases. Parity checking will only detect single bit errors. The reason it will only detect single bit errors is that if I was supposed to have for example six 1's, and two zeros in a byte but I lost two bits and now they look like zeros when they were supposed to be 1's, well I still have four and that's still even. So it wouldn't be recognized by parity checking. Also, parity checking provides no correction. So it's a very rudimentary method of error checking for memory. Error Correction Code came along a lot later than parity checking, and it's a much more sophisticated way of checking for memory errors, and it can detect and fix memory errors on the fly. Error Correction Code should be used for network Servers. Error checking by parity can be used for workstations, but doesn't necessarily have to be. But error checking is extremely important in regards to Servers, because servers are working on so many things at one time. And because the amount of memory and the amount of data transfer is tremendously higher in Servers. As with form factors, error checking is a component of the board or the Server itself - in other words, it comes with the system that we buy. Whether we can use error checking memory with that particular system will be determined by the board or the Server that we buy, and most servers will have this capability. There are ways that manufacturers can make available the capability to use more memory: this is by buffering or registering. And we will discuss buffering and registering in our next section.