Unraid (1 parity)
Unraid (1 parity)
Drive-independent parity. Each drive has its own filesystem; on failure only that drive's data is at risk. Mixed-size native, expandable one drive at a time.
How it works
Unraid does NOT stripe. Each data drive holds a complete XFS, BTRFS, or ZFS filesystem with its own files. A dedicated parity drive stores XOR parity calculated from the data drives' raw blocks. The parity drive must be ≥ the largest data drive. A drive failure is rebuilt onto a replacement; meanwhile the other drives stay readable directly with their own filesystems intact.
Formula: sum(data) — parity drive ≥ largest
Pros / Cons
Pros
- Each drive is independent — single failure = single drive lost (others still readable)
- Mixed drive sizes natural
- Add drives one at a time
- Simple recovery — drives still mount on any Linux box if you have to
Cons
- Paid license ($49+)
- No striping = single-disk read speed
- Parity drive must be ≥ largest data drive
- Parity-write penalty on every write (until offloaded to cache)
- Closed-source core
When to use
Media servers (Plex/Jellyfin), mixed drive sizes, gradual expansion, Docker / VM ecosystem. The homelab favourite.
When NOT to use
High-IOPS workloads — use RAID 10 / ZFS mirrors. Free / open-source-only requirements — use TrueNAS or OMV.
Rebuild math example
8 × 8 TB Unraid 1P (1 parity + 7 data). Failed data drive rebuild reads remaining 7 drives × 8 TB = 56 TB at 70 MB/s ≈ 110 hours. Long, but only that one drive's data is at risk during rebuild.