RAID stands for Redundant Array of Inexpensive Disks.
RAID levels that exists are levels 0 – 7, but the most popular are
0,1,3 and 5. I will give a brief overview of those levels below.
RAID provides you with redundancy and fault tolerance of your data,
in most configurations. This allows for high availability of your
data even when a hard drive completely fails.
Always have hotspares configured. A hotspare is a disk that is
configured to take the place of a failed disk. When a disk
failure in the RAID occurs, the hotspare will add itself to the array
to take the place of the failed disk. The RAID will then begin to
rebuild itself, at the same time still providing you access to your
data, naturally with a performance hit until the RAID has completed
rebuilding itself.
Use hardware RAID whenever possible. This means using a
controller, just like a SCSI controller, to configure and host the RAID.
- Higher performance because the hardware handles the array.
- You can boot from a RAID. In order to house your operating system on a RAID 1, you must use hardware RAID controllers.
- Greater features and flexibility.
- You can move the controller and hard disks to a different machine and your RAID is still intact.
Avoid software RAID configurations if possilbe.
- You take a major performance hit.
- You cannot boot to a software RAID because the operating system must be running in order to access the RAID.
- You cannot configure hotspares in a software RAID.
- You cannot move the RAID to a different machine. Only the
operating system that configured the RAID has the information required
to access and configure the array.
- If your OS crashes, so does your RAID. All gone, bye bye data.
RAID 0
- No redundancy in this configuration, but allows for the best
performance because of it. Fastest write performance of any RAID
configuration.
- At least 2 hard disks are required, usually of the same size and speed.
- Size is equal to the smallest hard drive in the RAID * the number of drives.
RAID 1
- Also referred to as “mirroring”. Redundancy is is acheived by
writing identical data to both drives. If one drive fails, the
other drives has all your data.
- Requires 2 and only 2 hard disks.
- The size of the array is the size of the smallest drive in the array.
- Commonly used for operating systems.
RAID 3
- In an R3 configuration, data gets striped across multiple disks at
the byte level. A single disk is used to store parity information
and the failure of any one disk, including the parity disk, will not
cause failure of the array. The major bottleneck in this solution
is the single parity disk, which much be accessed every time
information is written to the array.
- Requires at least 3 hard disks. 2 for storing data and 1 parity drive.
- Size of the array is equal to the size of the smallest drive * (total number of drives – 1)
- RAID 4 – Same as RAID 3, but the data is stiped in block and not bytes. This improves the performance of reading data.
RAID 5
- This is the most popular RAID configuration for storing data, such
as your database files. Unlike RAID 3, parity information is
striped across all disks in the array, just like the rest of the
data. This gives R5 a performance increase over R3
when writing data. Like RAID 4, it uses block level striping
for improved read performance. Fault tolerance is maintained by
making sure that the parity information for any particular block of
data is not written on the same drive that contains the actual data.
- Perfect for storing of data where the majorty of access to the RAID
will be read access, especially when configured with smaller stripe
sizes.
- Requires at least 3 hard disks.
- Size of the array is equal to the size of the smallest drive * (total number of drives – 1)
- RAID 6 – same as RAID 5, but parity information is
written twice. The difference in R5 and R6 is improved fault
tolerance, but with a performance hit.