Disk and RAID capacity calculator for home and small-office NAS
Type your drive sizes, compare every RAID level side by side in one table, then keep the level that fits your budget and your risk appetite.
🔒 100% client-side · zero server upload
Copies the plain-text table into the output box. Nothing is uploaded.
| Level | Usable TB | Usable TiB | Drives that may fail | Wasted TB |
|---|
Selected level
Filesystem overhead, and why
Every factor below is an estimate of what the filesystem keeps for itself, applied to the array’s usable capacity — never to the raw capacity, so parity is not counted twice.
| Filesystem | Set aside | Why that number |
|---|---|---|
| ext4 | 6.5 % · estimate | Reserves 5 % for root by default (“mke2fs -m 0” removes it) plus ~1.5 % for inodes and bitmaps. |
| XFS | 0.1 % · estimate | No reserved blocks at all; about 0.1 % goes to allocation metadata. |
| Btrfs | 3.5 % · estimate | Duplicates metadata by default and pads chunk/block groups: 2–5 %, so 3.5 % here. |
| ZFS | 4.2 % · estimate | Keeps 3.2 % slop it cannot allocate, plus RAIDZ/ashift padding and the uberblocks: ~4.2 %. |
| NTFS | 1 % · estimate | The MFT and its mirror, about 1 % on a multi-TB volume; it grows with the file count. |
| exFAT | 0.5 % · estimate | Two FATs and cluster slack, about 0.5 %; no permissions and no journal. |
| APFS | 3 % · estimate | Container metadata, snapshots and copy-on-write block sharing: about 3 %. |
| FAT32 | 1 % · estimate | The FAT itself, about 1 % — and a hard 4 GB limit per file. |
RAID is not a backup. Redundancy survives a dead disk; it does not survive a deleted file, ransomware, or a NAS that burns. Keep a copy somewhere else.
Malaysian rules & edge cases
A drive sold as “4 TB” holds 1012 bytes, but Windows and DSM count in TiB (240 bytes), so the same drive reads 3.64 TiB. Both numbers are correct — they just use different units. One rule explains almost every nasty surprise:
- Classic RAID truncates every drive to the smallest drive. On 4+4+8 RAID 5 you get (3−1)×4 = 8 TB, and the upper 4 TB of the 8 TB drive is wasted, not parity.
- Synology SHR only uses the upper slice when enough drives share it. SHR-1 needs at least 2 drives at that size, SHR-2 needs at least 3. With 4+4+8 the upper 4 TB of the lone 8 TB drive is unusable — SHR cannot protect a slice on one drive.
- RAID 10 tolerates 1 drive guaranteed, not 2. Two failures survive only when the two dead drives sit in different mirror pairs; if both land in one pair, the array is gone.
- Rebuild time is a best-case floor. The figures assume a full-throttle rebuild on an idle system. Under normal office load, real rebuilds run 2–4× longer.
- Filesystem overhead comes off the top — and it is an estimate. ext4 reserves 5% for root by default (
mke2fs -m 0removes it), Btrfs duplicates metadata and pads its chunks, ZFS keeps 3.2% slop it cannot allocate, XFS keeps almost nothing. The tool prints the factor it applies for the filesystem you pick, and charges it on the array’s usable capacity — never on the raw capacity, so parity is not counted twice.
| Level | Minimum drives | Usable (equal drives) | May fail |
|---|---|---|---|
| JBOD / single | 1 | sum of all drives | 0 |
| RAID 0 | 2 | all drives | 0 |
| RAID 1 | 2 | 1 drive | all but 1 |
| RAID 5 | 3 | (n−1) drives | 1 |
| RAID 6 | 4 | (n−2) drives | 2 |
| RAID 10 | 4, even | half the drives | 1 guaranteed |
| Synology SHR | 2 | like RAID 5/1, tier by tier | 1 |
| Synology SHR-2 | 4 | like RAID 6, tier by tier | 2 |
Which filesystem? On a Linux or Synology NAS with a classic md RAID array, ext4 is the safe default. Btrfs and ZFS are worth it when you want checksums and snapshots, but they build their own RAID — do not stack them on top of RAID 5, pick one or the other. Windows wants NTFS, a USB stick wants exFAT, and FAT32 should not hold video: it refuses any file over 4 GB.
Nothing you type is uploaded: this page runs entirely in your browser.
Before and after
Before
4TB, 4TB, 8TB — three drives, mixed sizesAfter
level usable_TB usable_TiB drives_may_fail wasted_TB free_TB
single 16.00 14.55 0 0.00 16.00
raid0 12.00 10.91 0 4.00 12.00
raid1 4.00 3.64 2 0.00 4.00
raid5 8.00 7.28 1 4.00 8.00
shr 8.00 7.28 1 4.00 8.00How it works
- Build the disk list: every drive is a card, and the capacity menu sits inside that card — the summary follows every choice, so there is nothing to run. Pick a shelf size and press "Add disk" (or type a custom size in GB or TB), then drag the cards into the order your bays are in; the arrows under a card do the same thing from the keyboard. More drives than fit stay on one row that scrolls sideways.
- Read the comparison table: each row is one RAID level with its usable TB, usable TiB, how many drives may fail, and how much capacity is wasted. When the disks differ in size the page says which one limits the array.
- Tap a row to see the waterfall (Raw capacity → redundancy → unusable → array usable → filesystem overhead → free space) and the rebuild time. Choose a filesystem to see the factor it charges and why. Press "Copy result" to copy the whole table.
FAQ
Why does my 4 TB drive show only 3.64 TB in Windows?
Windows and DSM count in TiB (2^40 bytes) while manufacturers sell in decimal TB (10^12 bytes). 4 TB = 3.64 TiB, so both numbers describe the same bytes. This calculator shows both so you can plan for what the OS will actually display.
Does RAID 10 survive two failed drives?
Only sometimes. RAID 10 is guaranteed to survive 1 failed drive. Two failures are fine only when the two dead drives are in different mirror pairs; if both failures hit the same pair, the array is lost. That is why the table prints 1, not 2.
Why does SHR ignore part of my largest drive?
SHR slices drives by capacity tier and needs enough drives in each tier to protect it: at least 2 for SHR-1 and at least 3 for SHR-2. A lone 8 TB drive next to 4 TB drives gives SHR a single-drive tier it cannot protect, so the upper 4 TB stays unusable until you add a second drive of that size.
How long does a real rebuild take?
The tool assumes a full-throttle rebuild, which is a best-case floor. An 8 TB drive at 140 MB/s rebuilds in about 15.9 hours that way. Under office load — active writes, parity checks, a slow replacement — expect 2 to 4 times longer.
Do you upload my drive sizes anywhere?
No. The calculator is static JavaScript in your browser. Open the Network tab and you will see zero requests. Your drive configuration is ordinary hardware data, but the rule is still: nothing leaves your machine.