RAID

RAID, a redundant array of independent disks, maps logical blocks across several member devices. A RAID level defines the placement of data and redundancy. The array appears as one block device to higher layers, while the RAID implementation tracks member state and reconstructs missing data when the layout permits.

Writes enter the logical array and are divided into chunks. Striping places chunks on different members. Mirroring stores additional copies. Parity stores reconstruction information calculated from data. Reads and writes are translated according to that layout. When a member fails, the array can operate in a degraded state by reading mirrors or reconstructing data from parity until a replacement is rebuilt.

RAID levels express layouts

RAID zero stripes data and has no redundancy. RAID one mirrors copies. RAID five stripes data with one distributed parity block per stripe. RAID six uses two parity blocks. RAID ten stripes across mirrored pairs. Names alone do not describe controller cache, chunk size, implementation, or failure domains, so those details still need documentation.