RAID 2 (or Redundant Array of Independent Disks, level 2) is one of many different schemes of mirroring and striping computer data drives in order to ensure that a minimal amount of data is lost. The other RAID levels are RAID 0, RAID 1, RAID 3, RAID 4, RAID 5, RAID 6, and RAID 10; each level has a different degree of tradeoff between cost and performance.
RAID 2 is also known as memory-style error-correcting-code (ECC) organization. Memory systems have long implemented error detection using parity bits, which helps to ensure that each byte is correct. Each byte in a memory system has a parity bit associated with it that records whether or not an even number of bits in that byte are set to 1 (even parity) or that an odd number of bits in that byte are set to 1 (odd parity). Even parity is usually designated by setting the parity bit to 0 and odd parity is usually designated by setting the bit to 1.
What good is this? To check that each byte is correct, all you have to do is count the number of 1s in the byte and see if it matches up with the parity bit. If it does, you're good; if it doesn't, then you have to perform some sort of error correction. There are various error correction schemes, but most require two to four additional bits for each byte. Thus, with a parity scheme like this, RAID 2 offers a degree of data integrity unavailable to the other RAIDs.
This concept of parity bits can be used in data storage as well. We designate eight data drives for data storage, and three to five more (depending on the error correction scheme) for parity bit and error correction bit storage. On the eight main storage drives, we place one of the bits in each byte, placing the first bit on drive 1, the second bit on drive 2, the third bit on drive 3, and so on. In this scheme, if a drive fails or causes an error, the parity bit can usually determine what the missing bit was; if there are multiple failures, the error correction scheme can help, too.
A vital consideration in a RAID 2 setup is the idea of homogenity among the various drives. This system only works as well as the smallest drive; you can only store eight times that much data, because one bit from each byte must fit on that smallest drive. If all of the drives have identical sizes and setups, then no space is wasted and the drives can be moved around quickly and efficiently. This also means that the possibility of having a single hot spare as a backup drive for the RAID 2 setup is quite feasible; a hot spare is a ready-to-go drive without any data that can be plugged in quickly and immediately be integrated into the system. Without homogenity, having hot spares available becomes considerably more expensive and logistically challenging. Essentially, RAID 2 nearly requires homogenity among the drives.
Another important note is that RAID 2 does not replace the concept of backups of critical data. The scheme provides great reliability on the go, but in the event of damage to all of the drives in a natural disaster or a power problem, a backup is essential. RAID 2, like the other RAID setups, is not a replacement for a backup scheme; it is merely insurance that your data will be available when you are actually using it.
RAID 2 requires eleven to thirteen drives to support eight drives worth of data; other schemes, such as RAID 1, require at least sixteen drives to store eight drives worth of data. In terms of pure storage, RAID 2 is quite good. However, the cost benefit in terms of the number of drives is lost when you consider the speed of using the system. Nine drives will constantly need to feed data, and this will require some care and expense when setting up the system so that all drives can deliver data with speed; likely, some sort of parallel delivery system will have to be implemented to have any sort of speed comparable to other RAID levels.
RAID 2 is an excellent scheme if you will not be requesting data in large amounts or extremely frequently, but when you need the data, it has to be there. In this case, RAID 2 is an excellent redundant scheme that saves noticeably on disk cost. An example of this is a situation where data is accessed on a regular basis and speed isn't vital, such as timesheet storage and retrieval. RAID 2 saves on cost and provides good data security, but loses out in the speed race without giving up the cost advantage.