I previously covered an article on the new research on RAID arrays, which has originated from FASTA Universidad. The latter half of their paper covers the different RAID levels that you should be aware of. Here is an overview of this very important research.
Learning the Nuances of the Seven Levels of RAID Arrays
The RAID specification is defined at seven different RAID levels. The seven RAID levels are as follows:
? RAID 0: RAID array disks are fragmented into chunks according to
bands, also called stripes, which allows the data handler to use the disks in parallel to
perform read/write operations on each disc. This RAID level provides
excellent performance, but no data redundancy.
? RAID 1: also called “mirror array”, defines a data disk and one or more disks.
Mirror disks are identical copies of the data disk and can be
automatically replaced if a fault occurs. This RAID level allows parallel reading
of the mirrored parts, but the main advantage of this way is that it
performs an “automatic backup” of the data disk to protect the information
against a physical rupture of all but one of the disks that make up the array.
? RAID 5: The disks in the RAID array are fragmented into bands, allowing the data handler to perform
parallel reading and writing on multiple disks. Each band
that corresponds to each disk is called chunk and an XOR function is calculated from the
N-1 chunks of band data. The parity chunk is distributed between disks for
different bands of the array, and thus prevents a single disc from loading with the array.
The writing task is associated with parity storage. This allows for better
performance of the entire array, as the use of the disks tends to be uniform.
Parity chunks allow the array to withstand the failure of one of the disks and continue
running in degraded form, waiting for the disk to change, and reconstruct the
lost disk when a replacement disk is available. RAID 5 is a balance between
capacity, performance and data redundancy.
? RAID 6: can be considered a variant of RAID 5 where two RAIDs are calculated.
Different parity functions allow the data handler to recover the array even if two
Disks fail. Again, there is a delicate balance between capacity, performance and redundancy.
In particular, this level of RAID makes it possible to retrieve from the RAID.
There may be a case where a second disc is broken during reconstruction.
In levels 5 and 6 it is also necessary to specify how the chunks will be distributed.
There is also a parity between the different bands. Three standard algorithms are usually applied, as shown here: Left
Asymmetric, Left Symmetric and Right Asymmetric. The difference between the three lies in the mode
in which parity and order are distributed.
RAID can be implemented by both software and hardware. The implementations of
Software RAID allow flexibility and provide different advantages at different levels. There is usually a tradeoff with performance, since the tasks required to manage the disks and
parity calculation must be carried out by the operating system, introducing the processor
as part of the critical path of the storage system.
In the case of a hardware-implemented RAID, some degree of flexibility is lost.
Performance is gained: the operating system delegates the tasks of
administration and calculation to the RAID controller. The controller is exposed to the host as one or more virtual disks.
In addition, RAID controllers can be implemented to write caches and circuits.
This requires a specific calculation for the parity functions, which allow to bring the arrangement closer to the theoretical maximum performance.
