Originally posted by DavPeanut
What is RAID exactly
A RAID is a Redundant Array of Inexpensive Disks. It entails using a hardware controller or a software driver to combine multiple hard drives together into one volume. There are different "levels" of RAIDs, which are really very different ways of combining disks.
Level 0 is "striping," where all the drives are combined together into one big volume. The advantage is blistering speed. If the computer needs to write 100 MB to a 4-disk RAID, it only has to write 25MB to each disk, so it goes four times as fast. Add more disks and it gets faster. The killer benchmarks on the Xserve are based on a 4-disk Level 0 RAID. However, the BIG disadvantage of striping is that if any drive fails, all the data is lost. In the case of that 100 MB file, 25MB worth of it would be gone, and there's no way to recover it. Striping is most often used for video cvapture, where they ghave to suck down a ton of data as fast as possible. Once the data is captured, then it is moved to a more reliable drive.
Level 1 is kinda the opposite. It is called mirroring. It only works with two drives, and the computer writes exactly the same information to each drive. There is no speed increase at all, but if one drive fails, NO data is lost at all. This provides extreme data security.
Level 0+1 mixes the two methods, and only works with 4 drives, I think. It is a mirror of two stripes. Or a stripe of two mirrors. In any event, data gets striped over two mirrors, so with four drives, it goes twice as fast as one drive, but there is no chance of data being lost.
Level 5 is the most popular one for big operations. It is more complex. It uses 3 or more drives (often up to 12, and even more). If n is the number of drives, it writes 1/(n-1) of the data to each drive except for one. So if there are 12 drives, it writes 1/11th of the data to each of 11 drives. The last drive is called a "parity drive," and it contains nothing but parity data. If one of those 11 drives fails, it can be rebuilt from the information on the other 11, plus the parity information. The L5 RAID can figure out what is missing by comparing the parity information to what information it still has. The advantages of L5 are extreme speed (with twelve drives, you supposedly go 11 times faster than one drive) and good protection against failure. The probems are the time it takes to rebuild a bad drive, and if two drives should happen to fail, all is lost. Also, the more drives you add the faster it is supposed to go, but the parity data gets more complex and there is more demand on the RAID controller, so it can start to slow down too.
Level 10 is a lot like 0+1, but it has no limitation on the number of drives. It is a mirror of two big stripes. So with 12 drives, you would have two striped arrays of 6 drives, and those two stripes would mirror each other. The disadvantage is that it uses a lot of drives compared to the amount of storage space you get. But the advantage is that it is EXTREMELY difficult to ever lose any data with L10, and it is still very fast. Also, there is no time spent creating parity data. So with a 12-drive L10 RAID, you get 6-drives worth of data writing at a full 6 times faster than a normal drive, and there is very very little chance of losing any data.
That's the brief overview.
Or at least my understanding of it. Do a search on "What is RAID" or "RAID levels explained" or something, and they'll probably give you better and more accurate info. Also, there are a lot of other levels of RAIDs, but from what I can tell, the vast majority only use the above levels.
