Popular Posts

Friday, January 31, 2014

Definition of:RAID (Redundant Array of Independent Disks)

A disk subsystem that increases performance or provides fault tolerance or both. RAID uses two or more regular hard drives and a RAID controller, which is plugged into motherboards that do not have RAID circuits. Today, most motherboards have built-in RAID but not necessarily every RAID configuration (see below). In the past, RAID has also been accomplished by software only but was much slower. In the late 1980s, the "I" in RAID stood for "inexpensive" but was later changed to "independent."

In large storage area networks (SANs), floor-standing RAID units are common with terabytes of storage and huge amounts of cache memory. RAID is also used in desktop computers by gamers for speed and by business users for reliability. Following are the various RAID configurations.



RAID 0
A popular disk subsystem that increases performance by interleaving data across two or more drives. Data are broken into blocks, called "stripes," and alternately written to two or more drives simultaneously to increase speed. For example, stripe 1 is written to drive 1 at the same time stripe 2 is written to drive 2. Then stripes 3 and 4 are written to drives 1 and 2 simultaneously and so on. When reading, stripes 1 and 2 are read simultaneously; then stripes 3 and 4 and so on.
 Ironically, RAID 0 is a misnomer because there is nothing "redundant." If one drive fails, the entire RAID array fails.



Disk Striping for Performance
Widely used for gaming, disk striping interleaves data across multiple drives for performance. However, there are no safeguards against failure.

RAID 0 Probability of Failure
The more drives in a RAID 0 array, the higher the probability of array failure. For example, if experience tells us that one out of a thousand drives fails in a year, the probability that a 2-drive array will fail in a year is 1 in 500; that a 3-drive array will fail is 1 in 333 and so on.

The formula: if the probability of failure of each of n drives is p for a given period, then the probability of not failing is (1-p). The probability of all drives functioning is (1-p)^n, and the probability that at least one drive will fail, causing the array to fail, is [1-(1-p)^n].

RAID 1
A popular disk subsystem that increases safety by writing the same data on two drives. Called "mirroring," RAID 1 does not increase performance. However, if one drive fails, the second drive is used, and the failed drive is manually replaced. After replacement, the RAID controller duplicates the contents of the working drive onto the new one.




Mirroring for Fault Tolerance
Widely used, mirroring writes two drives at the same time so that data are duplicated. It provides the highest reliability, but doubles the number of drives needed.

RAID 1 Probability of Failure
The more drives in a RAID 1 array, the lower the probability of failure. For example, if experience tells us that one out of a thousand drives fails in a year, the probability that an entire 2-drive array will fail in a year is 1 in a million; that an entire 3-drive array will fail is 1 in a billion and so on.

The formula: if the probability of failure of each of n drives is p, then the probability that all the drives will fail is p^n.

RAID 3

A disk subsystem that increases safety by computing parity data and increasing speed by interleaving data across two or more drives (striping). RAID 3 achieves the highest data transfer rate because all drives operate in parallel. Using byte level striping, parity bits are stored on separate, dedicated drives. Somewhat similar, RAID 4 uses block level striping but is not as popular.

RAID 3 - Speed and Fault Tolerance
With data striped across two or more drives, RAID 3 achieves the highest data transfer rate because all drives operate in parallel.
RAID 5
A popular disk subsystem that increases safety by computing parity data and increasing speed by interleaving data across three or more drives (striping). RAID 5 is similar to RAID 3, except that RAID 5 parity is distributed among all drives, whereas RAID 3 uses separate parity drives.

RAID 6
Similar to RAID 5 but not as widely used, RAID 6 performs either two parity computations instead of one, or it performs the same parity computation on overlapping subsets of the data. Using four drives, RAID 6 can recover from two failed disks.


RAID 5 - Speed and Fault Tolerance
With data and parity striped across three or more drives, RAID 5 has been a popular method for obtaining speed and safety.
RAID 10
 
A RAID subsystem that increases safety by writing the same data on two drives (mirroring), while increasing speed by interleaving data across two or more mirrored "virtual" drives (striping). RAID 10 provides the most security and speed but uses more drives than the more common RAID 5 method. S

RAID 10 Architectures
With RAID 1+0, the drives are set up as mirrored pairs and then striped. In this configuration, a drive in each mirrored pair can fail. In RAID 0+1, they are striped first, and then mirrored. If a drive fails in one of the stripe sets, then the other stripe set must be used, and no drive can fail in that set.

Thanks PCM


No comments:

Post a Comment