Hardware · Virtualization
RAID
Also known as: Redundant Array of Independent Disks
Combines multiple drives into a single logical volume for redundancy, performance, or both. The RAID level determines how data is distributed and how many drives can fail before data is lost.
RAID uses multiple physical drives together, presenting them as one volume to the operating system. Depending on the level, RAID can protect against drive failures, increase read/write throughput, or both — at the cost of some usable capacity.
RAID 0 — Striping, no redundancy
Data is striped across all drives in equal chunks. A read or write accesses multiple drives simultaneously, delivering the highest throughput of any RAID level. Usable capacity equals the sum of all drives.
The catch: zero redundancy. One drive failure loses everything. RAID 0 is appropriate only for scratch space, caches, or workloads where data loss is acceptable and performance is critical.
RAID 1 — Mirroring
Every drive has an exact mirror. Reads can come from either drive; writes go to both. Usable capacity is 50% — two 8TB drives give you 8TB usable. A single drive failure is fully tolerated; the mirror continues serving data.
RAID 1 is simple, reliable, and has excellent read performance. Write speed is limited to the slower of the two drives. Best for OS volumes, critical databases, and situations where simplicity and reliability matter more than capacity efficiency.
RAID 5 — Striping with distributed parity
Data and parity information are striped across three or more drives. Parity allows any single drive to be reconstructed if it fails. Usable capacity is n−1 drives — three 8TB drives gives 16TB usable.
RAID 5 balances capacity, performance, and redundancy. Read performance is excellent; write performance is lower due to parity calculation. Rebuild times after a drive failure are long on large drives, and during rebuild a second failure would be catastrophic. For anything mission-critical, RAID 6 is the safer choice.
RAID 6 — Striping with double distributed parity
Like RAID 5, but with two independent parity blocks distributed across the drives. Any two drives can fail simultaneously without data loss. Usable capacity is n−2 drives — four 8TB drives gives 16TB usable.
RAID 6 is the right choice for production storage that matters. The write overhead is slightly higher than RAID 5, but the additional protection against dual-drive failure is worth it — especially given that drives in the same array often fail in close succession (the stress of one failure and the subsequent long rebuild creates conditions for a second). For NAS arrays with large drives, RAID 6 is our standard recommendation.
RAID 10 — Mirroring + striping
RAID 10 (sometimes written RAID 1+0) combines RAID 1 mirroring with RAID 0 striping. Data is mirrored across pairs of drives, and those mirror pairs are then striped. Minimum four drives; usable capacity is 50% of total.
RAID 10 delivers the highest read performance of any redundant RAID level, fast rebuilds (only the mirrored pair needs rebuilding, not the full array), and can survive multiple drive failures as long as no mirror pair loses both drives. It's the choice for high-performance, high-reliability workloads: busy databases, VM datastores, high-throughput storage. The cost is efficiency — you're paying for half your raw capacity.
Choosing a RAID level
| Level | Min drives | Usable capacity | Drive failures tolerated | Best for | |---|---|---|---|---| | RAID 0 | 2 | 100% | 0 | Scratch space, caches | | RAID 1 | 2 | 50% | 1 | OS volumes, critical small datasets | | RAID 5 | 3 | n−1 | 1 | General file storage, moderate write loads | | RAID 6 | 4 | n−2 | 2 | Production NAS, large drives, critical data | | RAID 10 | 4 | 50% | 1 per mirror pair | Databases, high-performance VM storage |
RAID is not a backup. A RAID array protects against drive failure; it does not protect against accidental deletion, file corruption, ransomware, or catastrophic hardware failure. Maintain separate backups regardless of RAID level.