RAID?!
Arguably the most fascinating thing I learned this week was about RAID, the redundant array of inexpensive/independent disks. RAID uses more than one hard disk drive to store data, but uses different levels to determine the implementation. RAID level 0, for example, simply creates stripes of data across the disks. This means that a single file can be broke up and spread across all the disks. This creates the potential for faster file access because different parts of the file can be accessed at the same time (in parallel). Unfortunately, RAID 0 does not backup any data, and a loss of one disk can result in major data corruption since files are spread across disks.
The next level is RAID 1, which implements mirroring. This implementation simply writes the same data to two different disks. This way if a disk fails, there is a mirrored disk that can be used to back it up. Similar to RAID 1 is RAID 10 (or 1-0) which combines the striping aspect of RAID 0 with the mirroring of RAID 1. The main downside of RAID 1 (and by extension, 10) is that only half of available disk capacity can be used to store data (because the other half is backup).
RAID 4 attempts to solve the capacity problem by using a concept known as parity. Parity is a bit complicated, but essentially involves running a XOR function across data blocks and storing the result in a corresponding parity block. If any disk fails, it can be reconstructed by running the XOR function over the data blocks of the surviving disks (including the parity disk). If the failed disk is the parity disk, a new parity disk can be easily created in the same manner it was created originally.
RAID 5 is essentially the same as RAID 4, but the parity blocks are spread across all disks rather than being stored in a single parity disk (this is done for efficiency reasons). Although RAID 4 and 5 allow for N-1 max disk capacity usage, where N is the number of hard disk drives (as opposed to N/2 as seen with RAID 1 and 10), they only allow for failure of a single disk. RAID 6 mirrors the parity information and can therefore support failure of more than one disk.
RAID 4 attempts to solve the capacity problem by using a concept known as parity. Parity is a bit complicated, but essentially involves running a XOR function across data blocks and storing the result in a corresponding parity block. If any disk fails, it can be reconstructed by running the XOR function over the data blocks of the surviving disks (including the parity disk). If the failed disk is the parity disk, a new parity disk can be easily created in the same manner it was created originally.
RAID 5 is essentially the same as RAID 4, but the parity blocks are spread across all disks rather than being stored in a single parity disk (this is done for efficiency reasons). Although RAID 4 and 5 allow for N-1 max disk capacity usage, where N is the number of hard disk drives (as opposed to N/2 as seen with RAID 1 and 10), they only allow for failure of a single disk. RAID 6 mirrors the parity information and can therefore support failure of more than one disk.
No comments:
Post a Comment