Why RAID Exists
Hard drives fail. The mean time between failures (MTBF) for consumer hard drives is typically 1–3 million hours — impressive sounding, but a cabinet of 10 drives collectively reaches a failure event far sooner. RAID was developed to allow storage systems to keep serving data even when individual drives fail, without requiring immediate replacement and recovery from backup.
RAID is not a backup. This distinction is critical. RAID protects against drive failure — a hardware event. It does not protect against accidental deletion, ransomware encryption, filesystem corruption that propagates across all mirrors, or catastrophic events (fire, flood, theft) that destroy the entire array. The correct approach is RAID for availability plus a separate backup strategy for data protection.
How RAID Parity Works
RAID levels that survive drive failures use either mirroring or parity. Mirroring (RAID 1) stores identical data on two or more drives. If one fails, the mirror continues serving data unchanged. Simple but uses 50% of raw capacity for overhead.
Parity (RAID 5, 6) distributes a mathematical checksum across all drives. If one drive fails, the missing data can be reconstructed from the remaining drives and their parity information. RAID 5 distributes parity across N drives and can survive 1 drive failure with only 1/N capacity overhead (a 4-drive RAID 5 uses 3/4 of raw capacity, vs 1/2 for RAID 1). RAID 6 uses dual parity and can survive 2 simultaneous drive failures.
Hardware RAID vs Software RAID
Hardware RAID uses a dedicated RAID controller card with its own processor and cache. The controller presents the array to the OS as a single disk — the OS has no knowledge of the underlying RAID structure. Hardware RAID with battery-backed cache delivers excellent write performance for enterprise workloads. The downside: if the controller fails, you typically need an identical replacement to read the array.
Software RAID uses the CPU and OS to manage the RAID logic. Linux mdadm is the standard Linux software RAID tool. ZFS and Btrfs implement their own software RAID equivalents (RAIDZ, Btrfs RAID). Software RAID arrays are portable — move the drives to any machine and reconstruct the array. TrueNAS and Synology NAS devices use software RAID (ZFS-based on TrueNAS, mdadm or Btrfs on Synology) for this portability reason.
RAID Levels Compared
| RAID Level | Min Drives | Fault Tolerance | Usable Capacity | Read Speed | Write Speed | Best For |
|---|---|---|---|---|---|---|
| RAID 0 | 2 | None (1 failure = total loss) | 100% of all drives | Very fast (striped) | Very fast | Performance only; no redundancy needed |
| RAID 1 | 2 | 1 drive failure | 50% of total | Good (read from either) | Write speed of 1 drive | Small 2-bay NAS; boot drives |
| RAID 5 | 3 | 1 drive failure | (N-1)/N of total | Good (parallel reads) | Lower (parity calculation) | Balanced performance and capacity |
| RAID 6 | 4 | 2 simultaneous failures | (N-2)/N of total | Good | Lower (dual parity) | Larger arrays; high drive count |
| RAID 10 | 4 | 1 failure per mirror pair | 50% of total | Excellent | Excellent | Databases; high IOPS workloads |
| RAIDZ2 (ZFS) | 4 | 2 drive failures | (N-2)/N of total | Good | Good (ZFS ARC helps) | NAS with ZFS (TrueNAS) |
Frequently Asked Questions
Is RAID a backup?
No. RAID protects against drive failure only. It does not protect against accidental deletion (the deletion propagates to all mirrors immediately), ransomware (encryption propagates too), or disasters that destroy the whole device. Follow the 3-2-1 backup rule: 3 copies of data, on 2 different media types, with 1 copy offsite.
What RAID level should I use for a home NAS?
For a 2-bay NAS: RAID 1 (mirror) — simple, survives one drive failure, rebuilds quickly. For a 4-bay NAS: RAID 5 or RAIDZ1 gives good capacity efficiency with 1-drive fault tolerance; RAID 6 or RAIDZ2 gives 2-drive fault tolerance with slightly less capacity. For the highest data safety: RAIDZ2 on TrueNAS or SHR-2 on Synology.
How long does a RAID rebuild take?
A RAID 5 or 6 rebuild requires reading every block on every surviving drive to reconstruct the failed drive. For consumer hard drives at 150–200 MB/s, a 4TB drive takes roughly 6–8 hours to rebuild. During rebuild, the array is in degraded mode — another failure at this point causes total data loss in RAID 5. This is why RAID 6 (2-drive fault tolerance) is preferred for arrays with large drives or many drives.
What is RAID 0 used for?
RAID 0 (striping) splits data across multiple drives with no redundancy. Any single drive failure destroys all data. It is used purely for performance — read and write speeds scale linearly with the number of drives. Suitable for temporary scratch space, video editing working files, or any data you can afford to lose. Not suitable for storage you care about.