RAID is an acronym first defined in 1987 by researchers of the University of California, Berkeley as "Redundant Array of Inexpensive Disks". This storage technology allows a combination of multiple low-cost disk components into logical units (logical drives). A so called "RAID level" describes how data is distributed over the various disks of a logical drive. These distribution algorithms determine the balance of increasing data reliability through redundancy and/or increasing performance through parallel input/output transactions.
For marketing reasons the definition of RAID was later replaced by RAID manufacturers with the term "Redundant Array of Independent Disks". RAID technology should not be subject to a "low cost" expectation.
RAID is now a collective name for computer data storage schemes that can divide and replicate data among multiple disks.
The different RAID levels are named by the word RAID followed by a number, as in RAID-0, RAID-1, etc. A brief description of the best known RAID levels the controllers supported by ServerView RAID can configure follows:
A RAID-0 is created by distributing (striping) data over two or more disks. Single striping (like that just mentioned) does not generate any redundancy for protecting data, but by distributing the data over multiple drives it provides the best read/write performance of all RAID types.
A RAID-1 consists of two disks. The data stored on the array is written to both drives. The mirroring of data provides a redundancy which ensures that no data is lost if one drive fails. However, only half the total capacity of the two disks is available because all data is written to both drives.
In comparison with a single drive RAID-1 offers no benefits in terms of write performance, but because the data is distributed over two drives it provides advantages in read performance (and data security).
A RAID-1E consists of at least three drives. Some controllers require a configuration with an even drive count, e.g. MegaRAID SAS. Individual data blocks are mirrored onto the next disk (replication), and RAID-1E consequently offers better failsafe performance than RAID-1. However, neither two adjacent nor the first and last disks may fail simultaneously.
At least three drives are required to create a RAID-5. As with a RAID-0 the data is distributed over various drives, but in the case of RAID-5 the capacity of a drive is used to store parity information. The parity information is also distributed over all the drives. The controller generates this parity whenever data is written to the array and distributed over all the drives. If a drive fails, the content of the failed drive can be restored from the data and the parity of the remaining drives.
The use of parity minimizes the capacity costs of redundancy. As only one drive is used to store the parity, two thirds of the total capacity (in a 3 disk configuration) can still be used for data. In the case of arrays with more disks the reduction of the usable total capacity is less. With RAID-5 the write performance is lower because parity data must first be generated for each write process. The read performance is good, however, because the requests are distributed over all drives.
A RAID-6 requires at least four disks and functions in a similar way to a RAID-5, but can handle the failure of two disks. Some RAID controllers also offer a 3 disk configuration for RAID-6. Parity on 2 disks equals a double mirroring of data. In the case of a RAID-6 two pieces of parity information are calculated instead of one and these are distributed over all drives on a stripe-by-stripe basis. This is why a RAID-6 offers the highest level of security. Write access is a little slower than with RAID-5.
A RAID-10 is a dual-level array which is created by two or more equal-sized arrays of the type RAID-1 being used to produce a RAID-0. A top-level array (RAID-0) shares the total data load with the second-level array (RAID-1), thus enhancing both the read and the write performance. As second-level arrays are RAID-1s, redundancy is also offered. However, only half the total capacity of the drives used is available in the array.
A RAID-50 is a dual-level array which is created by using at least two arrays of the type RAID-5 to form a RAID-0. The top-level array (RAID-0) shares the data with the second-level array (RAID-5), thus enhancing both the read and the write performance. Since the second-level arrays use RAID-5, the parity provides efficient redundancy.
A RAID-60 is a dual-level array which is created by means of at least two arrays of the type RAID-6 in order to form a RAID-0. The uppermost array (RAID-0) shares the data with the array on the second level (RAID-6), which increases the read and write accesses. Using RAID-6 on the second level also guarantees a high degree of data security.
A single volume consists of a single disk. In the actual sense of the word this is not a real RAID type and is thus also referred to as a "none-RAID". According to the latest definition of the Storage Networking Industry Association a JBOD (Just a Bunch of Disks) is one of these although the term can sometimes refer to multiple physical disks.
A concatenation is formed by interconnecting two or more disks. In this case the drives can have different capacities and are interconnected from beginning to end. A global volume offers no redundancy and no performance benefits compared to a single drive; it is seen in the system merely as a correspondingly large drive.
A RAID volume is created by interconnecting two or more arrays of the same type. In contrast to the dual-level arrays described above, arrays in a RAID volume need not have the same capacity but are interconnected (as described above under concatenation).
Note: Sometimes the term "volume" is also used as a synonym for array.