Simple definition
RAID groups multiple disks into one logical storage system. Depending on its level, it can improve read/write performance, provide disk redundancy or both; RAID 0 has no fault tolerance.
Technical definition
RAID 0 stripes data without redundancy. RAID 1 mirrors data, RAID 5 uses distributed parity, RAID 6 uses dual parity, and RAID 10 combines mirroring with striping. These mechanisms help with certain disk failures but cannot replace an independent backup.
- Server / NAS Uses a logical volume
- RAID volume Presented storage
- RAID level 0, 1, 5, 6 or 10
- Physical disks Mirrors or parity
- Fault tolerance Depends on the level
RAID 0 provides no disk-failure tolerance. RAID 1, 5, 6 and 10 offer different redundancy levels; RAID is not a backup.
What is it used for ?
Support storage availability or performance on servers and NAS devices, according to the selected RAID level and required capacity.
Practical example
A properly configured RAID 6 array with at least four disks can tolerate the failure of up to two disks without losing the logical volume, but the disks still need to be replaced and the array rebuilt.
Common issues
- Mistaking RAID for an independent backup
- Long rebuild times and increased exposure while degraded
- Correlated failures affecting multiple disks
- Undocumented controller, disk or array configuration
Comparing RAID 0, 1, 5, 6 and 10
RAID 0 stripes data across disks and may improve throughput, but a single disk failure can destroy the logical volume. RAID 1 mirrors data onto multiple disks and protects against some drive failures at a cost in usable capacity. RAID 5 distributes data and parity and typically tolerates one drive failure. RAID 6 uses dual parity and can tolerate two failed drives. RAID 10 stripes mirrored groups; it can offer good performance, but its fault tolerance depends on which disks fail.
Usable capacity varies with the RAID level, disk count and disk size. Combining unequal-sized disks can reduce usable space depending on the controller and implementation.
What happens when an array is degraded?
A redundant RAID volume may remain online after a supported disk failure, but it is operating in a degraded state. Replacing the failed disk normally starts a rebuild that reads extensively from surviving drives. Large arrays or aging disks can require lengthy rebuilds, creating a period of increased exposure to additional failure.
Rebuilding is not the same as consistency checking, scrubbing or examining drive SMART information. These operations address different aspects of reliability. Controller alerts, NAS logs and the status of independent backups should inform recovery decisions.
Why RAID cannot replace a backup
Accidental deletion, ransomware encryption or application corruption may be reflected immediately across every disk in a RAID set. The entire system also remains vulnerable to theft, fire and controller faults. A snapshot kept on the same storage is not automatically an independent backup.
RAID primarily supports continued access during certain disk failures. Backup enables restoration of data from an earlier point in time. A resilient design may use both, with isolated backup copies and regularly tested recovery.
Related terms
Key takeaway: RAID level determines usable capacity, performance and fault tolerance; independent backups and recovery tests are still necessary.