RAID (Redundant Array of Independent Disks) combines several physical disks into one logical volume so that, depending on the level, the array survives a disk failure without losing data, or goes faster, or both. In surveillance it is the standard way to make recording storage resilient: a disk will eventually fail, and RAID is what stops that failure from taking the footage with it. The common levels trade usable capacity, performance, and how many disks can fail.
The levels each suit different needs. RAID 1 mirrors two disks; RAID 5 stripes data with one parity disk (survives one failure) at the cost of a write penalty; RAID 6 keeps two parity disks (survives two failures), important on large arrays; RAID 10 mirrors and stripes for speed plus redundancy. Surveillance is overwhelmingly write-heavy — cameras stream to disk continuously — so the write penalty of parity levels matters: a RAID 6 array writes each block with extra parity operations, which must be accounted for in the sustained throughput the array can take.
The defining pitfall is the rebuild window. When a disk fails and is replaced, the array rebuilds onto the new disk, which on large modern drives can take many hours or days, during which the array is degraded and a second failure (or an unrecoverable read error) can destroy a RAID 5 array entirely. This is why large surveillance arrays favour RAID 6 or 10 plus a hot spare and free headroom, not RAID 5 — and why RAID is resilience, not a backup: it does not protect against deletion, fire, or ransomware, which need a separate off-array copy.
