RAID
- Tags
- soft-eng
The employment of [see page 16, multiple] hard disks together to provide data redundancy (backups).
| Level | Description |
|---|---|
| RAID-0 | Data is stripped (segmented) across different disks. |
| RAID-1 | Mirror copy of data is also made across different disks. |
| RAID-2 | Error correcting codes also retained on different disks. |
For RAID-1 If one disk fails then we still have the same data available on another disk, until that data can be cloned to a new backup disk as well. Reads are easier in RAID than writes because you can READ from any one backup but must write to all the backup disks.
RAID-0 doesn't provide redundancy but it can fasten disk writes. Since part of a file could be written to one disk and another to another disk, you can 1 parallelise the writes to improve write times.
RAID-3 attaches parity bits for every N-bits written that ensures the total number of 1s written is even. Parity bit is saved to a different disk. If any disk including the parity disk is lost, then its contents can still be computed using the contents of the other disks. For example to you could have 3 backup disks and one parity disk. To store 6 bytes you store byte 1 in disk 1, byte 2 in disk 2, byte 3 in disk 3, a parity byte in a separate disk and then repeat for bytes 4-6. If any of disk 1, 2, or 3 fail then you can reconstruct the missing byte from the remaining disks and the parity disk. This RAID version has the maximum throughput since data is split amongst as many disks as possible and can all be read in parallel. However the main weakness is if more than one disk fails then there's no reliable way to reconstruct the lost data.