Mixed drive sizes
Say you've got two 12 TB drives and two 8 TB ones, 40 TB raw. RAID 5 treats all four as 8 TB drives, which leaves you 24 TB with 8 TB doing nothing at all. SHR is cleverer about it. It builds an 8 TB layer across all four drives and a 4 TB layer across the two big ones, and you end up with 28 TB. Unraid with one parity drive lands on the same 28 TB by a different route, with a 12 TB drive on parity and the other three counting in full.
With a drawer full of odd sizes, then, it's SHR on a Synology and Unraid on your own hardware. ZFS is the fussy one. Buy matching drives for it, or pair them up into mirrors of equal size.
TB or TiB?
Drive makers count a terabyte as a trillion bytes. Windows counts in tebibytes (240 bytes, nearly 10% bigger) and then labels them TB anyway, which is why a brand-new 16 TB drive turns up in Explorer as 14.55 TB and looks like it's been short-changed. macOS counts in decimal like the drive makers and shows the full 16 TB.
| On the box | In Windows |
|---|---|
| 4 TB | 3.64 TB |
| 8 TB | 7.28 TB |
| 12 TB | 10.91 TB |
| 16 TB | 14.55 TB |
| 20 TB | 18.19 TB |
| 24 TB | 21.83 TB |
How each layout counts space
- SHRSynology · 1+ drives
- Synology slices the drives into layers at each size and protects each layer on its own, and that's how the extra space on bigger drives gets used once two of them share it.
- SHR-2Synology · 4+ drives
- Same layering, two drives of protection. Needs four drives or more.
- RAID 0RAID · 2+ drives
- Adds every drive together and keeps nothing spare. One failure and the lot's gone.
- RAID 1RAID · 2+ drives
- Every drive holds the same data, and usable space is one drive (the smallest).
- RAID 5RAID · 3+ drives
- One drive's worth of parity spread over all of them. Usable is (drives − 1) × the smallest drive.
- RAID 6RAID · 4+ drives
- RAID 5 with two drives' worth of parity. Usable is (drives − 2) × the smallest.
- RAID 10RAID · 4+ drives
- Mirrored pairs striped together, giving you half the drives × the smallest.
- ZFS mirrorZFS · 2+ drives
- Two-way mirror vdevs in one pool, with each pair counting as its smaller drive.
- RAIDZ1ZFS · 3+ drives
- One drive of parity in a single vdev, sized to the smallest drive. ZFS's answer to RAID 5.
- RAIDZ2ZFS · 4+ drives
- Two drives of parity in one vdev, the ZFS take on RAID 6.
- RAIDZ3ZFS · 5+ drives
- Three drives of parity, for wide vdevs of big drives.
- Unraid, 1 parityUnraid · 2+ drives
- The biggest drive holds parity and every other drive adds its full size.
- Unraid, 2 parityUnraid · 3+ drives
- The two biggest drives hold parity. Everything else counts in full.
- SnapRAID, 1 parityDAS / PC · 2+ drives
- Counts like Unraid, with a parity drive at least as big as the largest data drive. Parity only updates when you run a sync, which suits a media library that rarely changes.
- SnapRAID, 2 parityDAS / PC · 3+ drives
- Two parity drives, synced on a schedule. Worth it on bigger DAS or PC arrays.
- JBOD, separate disksDAS / PC · 1+ drives
- Every drive on its own and no parity. A dead drive takes only its own files with it.
What the calculator leaves out
- File system overhead. Btrfs and ext4 keep a few per cent back for metadata.
- ZFS slop space, the 1/32 of the pool (up to 128 GiB) that OpenZFS holds back.
- Hot spares and SSD cache drives. Leave those bays out of the count.
- The small system partition Synology and QNAP put on every drive.