Author: Tom Large
Last Modified: 01-02-2K1
Synopsis: This is a quick synopsis of RAID, what it is and how it works. It is by no means a thorough in depth all-encompassing document, just a quick explanation of the basics. If you have any questions or comments, I can be reached at tlarge@thelug.org.
RAID, as it pertains to most Systems Administrators.
RAID stands for either Redundant Array of Independent Disks, or Redundundant Array of Inexpensive Disks, depending on who you ask. Basically, RAID is a grouping of disks pulled into a group, or an array. There are various levels of RAID, the more common ones will be described below.
Definitions:
Physical Disk: This can be a hard drive, a tape drive, a CD ROM, or any other mass storage device. It is a reference to the physical disk itself, regardless of its configuration.
Logical Disk: This can be a partition, a drive letter, or a directory depending on its implementation. A single physical disk can contain many logical disks. Also, a single logical disk can span many physical disks. A logical disk is the entity that your operating system "sees" as a place it can store data.
Parity: As defined by Webster's Dictionary: "the state of being odd or even used as the basis of a method of detecting errors in binary-coded data. What this means, in the world of storage, is a section of drive space that is used basically as an immediately available 1: way to verify that its corresponding data is in good condition, and 2: as a backup.
Stripe: In RAID sets, the term stripe means to spread data, typically in set increments (64k for example), across two or more physical disks.
RAID Levels:
RAID Level 0 (Disk Striping):
RAID Level 0 requires at least two physical disks, preferably three or more. RAID 0 is the act of striping data across many disks to gain better performance, but no redundancy. There is no parity in RAID 0. The advantage is speed. An Ultra Wide SCSI-2 bus for example can handle 40 MB per second throughput. Each physical disk can handle a sustained throughput of approximately 8 MB per second. If in our example we were to use five 20 GB physical disks, we would get the full 40 MB per second throughput to a single 100 GB logical drive, instead of 8 MB per second throughput to five individual 20 GB drives. Useable disk space can be calculated as ([Number of Disks] * [Size of Disks]).
RAID Level 1 (Mirroring):
RAID Level 1 requires two physical disks, preferably of equal size. RAID 1, also known as disk mirroring, is the process of using two disks, one for data and one for parity. Each disk is an exact duplicate of the other. The benefit here is that at any time, you have a complete fail-over solution to your disk. Useable disk space can be calculated as ([Number of Disks] / 2), meaning you get exactly half the disk space you pay for.
RAID Level 4 (Disk Striping with Parity):
RAID Level 4 requires at least three physical disks. With RAID 4, data is striped across all but one disk, with that disk used for parity. The one disk is used as a hot spare. RAID 4 can tolerate the loss of up to one disk, with no data loss, as the parity disk will kick in and take over for the lost drive. Useable disk space can be calculated as (([Number of Disks] - 1) * [Size of Disks]).
RAID Level 5 (Disk Striping with Parity):
RAID Level 5 requires at least three physical disks. RAID 5 operates very similar to RAID 4, except that the parity is spread across all disks, instead of just one. In general, this is the RAID configuration you will see in most corporate environments. Useable disk space can be calculated as (([Number of Disks] - 1) * [Size of Disks]).