A parity drive is a disk in a storage array dedicated to holding calculated recovery data that can rebuild the contents of any other drive in the group if it fails. Instead of storing a duplicate copy of your files (which would cost you half your total storage), parity uses a mathematical shortcut that requires only one extra drive’s worth of space to protect all the others.
How Parity Math Works
Parity relies on a simple binary operation called XOR (exclusive or). XOR compares bits across your data drives and produces a single result bit: it outputs a 1 whenever an odd number of input bits are 1, and a 0 when the count is even. That’s it. This one rule is enough to reconstruct any single missing piece.
Here’s a concrete example. Say you have three data drives storing the binary values 1, 0, and 1 at a given position. XOR those together: 1 ⊕ 0 ⊕ 1 = 0. The parity drive stores that 0. If the second drive dies, the system XORs the surviving values (1, 1) with the parity value (0): since the result must match, it calculates the missing bit as 0, which is exactly what was lost. This works no matter which drive fails, because XOR is reversible in every direction.
Scale that operation across trillions of bits and you have a parity drive: one disk that stores nothing but these calculated check values, yet can regenerate the full contents of any single failed neighbor.
Parity in RAID 5 and RAID 6
The most common place you’ll encounter parity drives is in RAID arrays. In RAID 5, parity data is distributed evenly across all drives in the array rather than sitting on one dedicated disk. This spreads the read and write load so no single drive becomes a bottleneck. The total parity overhead equals one full drive’s capacity. So a four-drive RAID 5 array with 4 TB drives gives you 12 TB of usable space, with the equivalent of one drive’s worth reserved for parity.
RAID 6 takes this a step further by writing two independent sets of parity data, costing you the equivalent of two drives’ capacity. The payoff is significant: a RAID 6 array survives two simultaneous drive failures. If corrupt parity data is found during a rebuild, the second copy steps in. For mission-critical data, that extra layer of protection matters more than the lost capacity.
What Happens When a Drive Fails
When a drive in a parity-protected array dies, the array enters a degraded state. It’s still functional, and you can still read and write files, but there’s no safety net until the failed drive is replaced. Once you swap in a new disk, the rebuild process starts automatically.
During a rebuild, the RAID controller reads every bit of data from all surviving drives, runs the XOR calculations against the stored parity, and reconstructs the missing data block by block. It writes the recovered data to the replacement drive, then re-synchronizes the full array. At that point, redundancy is restored and the system returns to normal operation.
Rebuild times depend on array size and drive speed. A few terabytes might finish in hours. Larger arrays can take a day or more, and during that entire window, the array is vulnerable. If a second drive fails before the rebuild completes in a RAID 5 setup, the data is gone.
The Rebuild Risk With Large Drives
This vulnerability gets worse as drive capacities grow, due to something called an unrecoverable read error (URE). Consumer drives are typically rated for one read error per 10^14 bits read, which works out to roughly one error per 12.5 TB of data. During a rebuild, the controller must read every bit on every surviving drive. With three 8 TB drives in a RAID 5 array, the total read during reconstruction exceeds that 12.5 TB threshold, making a URE more likely than not.
A URE during rebuild means the controller hits a spot on a surviving drive that it can’t read. In a RAID 5 array, that single error can make the rebuild fail, potentially taking your entire array down. This is the main reason storage professionals have shifted toward RAID 6 or other dual-parity solutions for arrays using large drives. Enterprise drives are rated for far fewer UREs (one per 10^15 or 10^16 bits), which helps, but the math still favors dual parity once individual drives exceed a few terabytes.
Software Parity: RAID-Z and Alternatives
Traditional hardware RAID controllers handle parity at the block level using fixed-size stripes across disks. Software-defined alternatives like ZFS’s RAID-Z take a different approach. RAID-Z uses dynamic stripe widths, adjusting the stripe size based on how many blocks are being written at any given moment. This eliminates a problem called the “write hole,” where a power failure during a partial stripe write can leave parity and data out of sync.
RAID-Z comes in three tiers: RAID-Z1 (single parity, like RAID 5), RAID-Z2 (double parity, like RAID 6), and RAID-Z3 (triple parity, no traditional equivalent). For home servers and NAS setups, software parity through ZFS or similar filesystems has become increasingly popular because it adds data integrity checks on top of the parity protection, catching silent corruption that hardware RAID would miss entirely.
Parity vs. Mirroring
The alternative to parity-based protection is mirroring (RAID 1), where every drive has a complete duplicate. Mirroring is simpler, faster to rebuild, and doesn’t carry the URE risk during reconstruction since there’s a full copy rather than a calculated one. The tradeoff is cost: mirroring cuts your usable storage in half.
Parity shines when you need to maximize usable capacity across many drives. In a ten-drive RAID 5 array, you lose only one drive’s worth of space to protection. The same ten drives in a mirrored setup would give you only five drives of usable space. That efficiency gap is why parity remains the standard approach for large storage arrays where capacity and fault tolerance both matter.