Redundant Array Inexpensive Discs

RAID

Most servers use Hard Disk Virtualisation to lump several hard disks together. Once that config has been decided and actioned, the new virtual disks can be connected/presented to the server.  The server will only see the virtual disk not the individual disks.

Thought I'd take a wee note and put down what RAID configs are the most common. Thanks to SCSI hardware and software we can slice n dice big trays of hard disks anyway we want to provide greater functional scope!!

RAID was originally called Redundant Array of Inexpensive Disks it now means Redundant Array of Independent Disks.

The main goals of RAID is to provide fault tolerance by redundancy and performance through disk striping.
Striping distributes data over several hard disks and spreads the load over the hardware. Redundancy is additional information that is stored so that operations can continue in the event of a disk failure.

A RAID controller writes the data to the virtual hard disk, the hardware and protocols will distribute the data automatically across the physical disks in various manners, these are called RAID levels.

RAID 0 = block by block striping
Distributes data written to the virtual disk on to one physical disk after another. So if disk A is busy next block is written to disk B then disk C...increases performance. I suppose this can be looked at as a sequential writing of data. Increases Performance

RAID 1 = block by block mirroring
Simplest example brings two hard disks together to form a virtual disk. Data is duplicated between the disks. Provides redundancy. Increases Fault Tolerance.

RAID 2 = Hamming Code
Defunct config. A technique used to detect disk data errors by storing redundancy data as well as real data. This functionality or comparative functionality is performed by disks and controllers now so the level is defunct.

RAID 3 = Parity Data
Stores parity data and stores data of a block amongst disks. All disks are involved in  read/writes. All disk rotations on the spindle are synchronised for simultaneous access. It was recommended for sequential write/read profiles. Now less important due to modern capabilities of hard disks and hard disk cache capabilities/techniques.

RAID 4 = PARITY
Idea of  PARITY is to replace mirrored disks with a parity disk.
One of the disk out of the group of disks will be kept as a parity disk. Data will be striped across the remaining disks. In the event of failure, the Parity disk and the Data disks will combine data/parity blocks to reconstruct the original data on the replaced disk.

RAID 5 = STRIPED PARITY
The Parity blocks are distributed sequentially across all the physical disks in the virtual disk. Now if a disk fails the remaining disks will use the data and parity blocks to rebuild the data.

RAID 0+1 = 2 stage Virtual disk hierarchy
Step 1 = Raid 0 = block by block striping. Step 2 = Raid 1 = block by block mirroring amongst several virtual Raid 0 disks. So we combine the two techniques! Apply it to several Virtual Disks

RAID 10 = 2 stage Virtual disk hierarchy
Kind of the reverse of RAID 0+1... bring the physical disks together using RAID 1 mirroring. Now consolidate the virtual disks with RAID 0 stripping. Again only the second level config is presented to the server.

In both RAID 0+1 and RAID 10 the server sees only a single disk which is faster, larger and more fault tolerant than the standard physical disks.

For a more in depth look at RAID Levels check this wikipedia page.
For a good book on Storage networks and RAID I recommend this book.