RAID Capacity Calculator: SHR vs ZFS vs RAID 5, 6 and 10
Enter your drives once and this calculator shows usable capacity, parity overhead, stranded capacity and fault tolerance for Synology SHR and SHR-2, ZFS RAIDZ1, RAIDZ2 and RAIDZ3, and standard RAID 0, 1, 5, 6 and 10 side by side. Four 8 TB drives give 24 TB usable under RAID 5, RAIDZ1 or SHR, and 16 TB under RAID 6, RAIDZ2 or RAID 10.
Every RAID calculator on the internet belongs to a hardware vendor, so each one models exactly one storage scheme. Synology's models SHR. TrueNAS-flavoured tools model RAIDZ. Controller vendors model the classic levels. No manufacturer tool will show you all three for the same physical drives, which is precisely the comparison you need while a shopping cart is open. That is what this page does.
Comma separated, mixed sizes welcome. Shorthand like 6x12 expands to six 12 TB drives. Use the size printed on the drive label.
| Layout | Usable TB | Usable TiB | Parity | Stranded | Survives | Rebuild reads | URE risk |
|---|
Rebuild reads is how much data the array must read to reconstruct one failed member. URE risk applies the manufacturer's unrecoverable read error rating to that read volume. Read the caveat below before treating it as a forecast.
The math, so you can check it
Every figure above comes from four short rules. None of them is proprietary, and none of them requires trusting this site.
- Standard RAID and ZFS truncate to the smallest drive. With n drives and p drives of parity, usable capacity is (n - p) x smallest. Everything above the smallest drive's size on every larger drive is stranded.
- RAID 1 is the exception in the other direction. An n-way mirror gives you one drive of capacity and survives n minus 1 failures.
- RAID 10 gives you half. Usable is (n / 2) x smallest. It is guaranteed to survive one failure and can survive up to n over 2 if you are lucky about which drives die.
- SHR slices instead of truncating. See the next section, because this is the part vendor tools disagree about.
How Synology SHR actually works
SHR is not a new parity algorithm. It is standard Linux software RAID arranged cleverly. Synology slices the drives horizontally at every distinct drive size, then gives each horizontal band its own parity group.
A band that sits on k drives yields (k - p) x band height of usable space when k > p. When k is not larger than p, that band has too few drives to protect and the whole band is stranded. Worked through on 2 TB, 3 TB and 4 TB drives with SHR:
| Band | Height | Drives in band | Result | Usable |
|---|---|---|---|---|
| 0 to 2 TB | 2 TB | 3 | Three-drive parity group | 4 TB |
| 2 to 3 TB | 1 TB | 2 | Two-drive mirror | 1 TB |
| 3 to 4 TB | 1 TB | 1 | Nothing to protect it, stranded | 0 TB |
| SHR total | - | - | - | 5 TB |
| RAID 5 total | - | - | Every drive truncated to 2 TB | 4 TB |
That 1 TB gap is the entire SHR pitch, and it grows as your drives diverge. With identical drives SHR collapses to plain RAID 5 and the advantage disappears completely, which is worth knowing before you pay a premium for it. The corollary matters more: SHR is the only consumer scheme that makes a rolling drive-upgrade path work. On ZFS or standard RAID, replacing two drives out of four with larger ones gains you nothing at all until every member is upgraded.
Whatever layout you land on, the drives are the purchase that matters. An enclosure is a one-time few hundred dollars. Four to eight drives is where the real money goes, and unlike the box they get replaced and upgraded over the life of the array. Buy CMR drives with a NAS workload rating, and see our NAS drive comparison before committing to a capacity.
TB against TiB, the source of nearly every "missing capacity" question
A drive maker sells you decimal terabytes. One TB is exactly 1,000,000,000,000 bytes. Your operating system reports binary tebibytes. One TiB is 2 to the power of 40 bytes, or 1,099,511,627,776. Divide one by the other and you get 0.909495, so a 20 TB drive genuinely holds 18.19 TiB.
Nothing was lost and nobody cheated you. Windows and most NAS interfaces just label a TiB with the letters "TB". Linux tools with -h flags usually mean TiB too. The practical effect is a fixed 9.05 percent apparent shrink at every capacity, before any filesystem overhead.
| Label on the drive | Bytes | Reported as | Apparent loss |
|---|---|---|---|
| 4 TB | 4.0 x 10^12 | 3.64 TiB | 9.05% |
| 8 TB | 8.0 x 10^12 | 7.28 TiB | 9.05% |
| 12 TB | 1.2 x 10^13 | 10.91 TiB | 9.05% |
| 16 TB | 1.6 x 10^13 | 14.55 TiB | 9.05% |
| 20 TB | 2.0 x 10^13 | 18.19 TiB | 9.05% |
| 24 TB | 2.4 x 10^13 | 21.83 TiB | 9.05% |
On top of that unit conversion, the filesystem takes a genuine cut. Ext4 and Btrfs reserve roughly one to two percent for metadata and superblock copies. ZFS reserves 1/64 of the pool, about 1.56 percent, as slop space, and separately you should plan to keep a ZFS pool below about 80 percent full because write performance degrades sharply above that. Synology DSM also carves a small system partition and swap area out of every drive. None of those are the reason your 20 TB drive says 18.2, but together they explain the last percent or two.
Fault tolerance is not the same as rebuild safety
A layout that "survives one drive failure" only survives it if the rebuild completes. That rebuild is the most stressful thing an array ever does, because reconstructing a failed member means reading every sector of every surviving drive, on drives that are the same age and often from the same manufacturing batch.
Two numbers matter here. The first is time: a 20 TB member resilvering at a realistic 80 to 150 MB per second takes roughly 37 to 70 hours, and longer if the array is also serving files. The second is read volume, shown in the table above. A six-drive RAID 5 of 12 TB drives has to read 60 TB to rebuild one member.
Manufacturers rate consumer drives at one unrecoverable read error per 10^14 bits and NAS or enterprise drives at one per 10^15. Treating those as independent events, 60 TB of reads against a 10^15 rating comes out around a 38 percent chance of encountering one.
Read that number carefully. The URE rating is a warranty floor, not a measurement, and observed error rates across large drive fleets run substantially better than spec. Modern systems also survive a URE rather than aborting: ZFS reports which file was affected and continues, and mdadm can be told to carry on. Use this figure to rank layouts against each other, not as a prediction that your rebuild will fail.
The honest conclusion is still the conventional one. Past roughly 8 to 10 TB per drive, use double parity. RAID 6, RAIDZ2 and SHR-2 all keep protecting the array during the window when it is most exposed. See our RAID level comparison chart for the full matrix, and the drive failure probability calculator to put annualised failure rates against your own drive count.
Which layout to actually pick
| Situation | Pick | Why |
|---|---|---|
| Two-bay NAS, first one | SHR or RAID 1 | Identical either way with matched drives, but SHR leaves the upgrade door open |
| Four bays, drives under 8 TB | SHR or RAID 5 or RAIDZ1 | Single parity is defensible while rebuild windows stay short |
| Four bays, drives 10 TB and up | SHR-2 or RAID 6 or RAIDZ2 | Rebuild takes over a day, and you want cover during it |
| Six or more bays | RAIDZ2 or SHR-2 | More members means more chances of a second failure mid-rebuild |
| Drives you already own, all different sizes | SHR | The only scheme that uses the extra capacity instead of stranding it |
| Virtual machine or database storage | RAID 10 | Rebuilds read only the mirror partner, and random write performance is far better |
| Scratch space you can lose | RAID 0 or JBOD | No redundancy, so only for data that exists somewhere else |
Related tools
- Drive failure probability calculator for the odds across your whole array over several years
- 3-2-1 backup sizing calculator because the array is only copy one
- Power draw calculator to see what all those spindles cost to keep spinning
- RAID levels explained for the plain-language version of everything above
- Usable capacity chart for common drive and bay counts, precalculated
Frequently asked questions
How much usable space do I get from four 8TB drives?
It depends entirely on the layout. RAID 5 and RAIDZ1 give 24 TB usable with one drive of parity, RAID 6 and RAIDZ2 give 16 TB with two, RAID 10 gives 16 TB, and Synology SHR gives 24 TB because all four drives are the same size. Your NAS will report those figures as 21.83 TiB, 14.55 TiB and 14.55 TiB respectively, which is the same capacity counted in binary units.
Why does my 20TB drive show up as 18.2TB?
Nothing is missing. Drive makers sell decimal terabytes where 1 TB is exactly one trillion bytes, while Windows and most NAS interfaces report binary tebibytes where 1 TiB is 1,099,511,627,776 bytes. Dividing gives a ratio of 0.909, so a 20 TB drive holds 18.19 TiB of the same bytes. Filesystem metadata then takes a further one to two percent, which is real but much smaller than the unit difference.
Is Synology SHR actually better than RAID 5?
With identical drives the two produce exactly the same usable capacity, because SHR falls back to a single RAID 5 group. SHR only wins with mixed drive sizes, where it slices the drives into bands and protects each band separately instead of truncating every drive to the smallest one. Three drives of 2 TB, 3 TB and 4 TB give 4 TB under RAID 5 and 5 TB under SHR, and the gap widens as your drives diverge.
Should I use RAID 5 or RAID 6 with large drives?
Use double parity once individual drives pass roughly 8 to 10 TB. The reason is rebuild exposure rather than a hard rule: a 20 TB member can take well over a day to resilver, the array reads every surviving drive during that window, and a second failure or an unreadable sector during a single-parity rebuild costs you the array. RAID 6, RAIDZ2 and SHR-2 all keep protecting you while the first rebuild runs.
Does RAID replace backups?
No, and this is the most expensive misunderstanding in home storage. RAID protects against a drive dying. It does nothing about accidental deletion, ransomware, a failed controller writing garbage to every member, theft, fire or flood, because every one of those events reaches all drives at once. Treat the array as your fast working copy and keep at least two more copies, one of them off site.
Can I mix drive sizes in the same array?
You can, but only Synology SHR and SHR-2 use the extra capacity. Every standard RAID level and every ZFS RAIDZ vdev truncates each member to the size of the smallest drive, so adding one 4 TB drive to a set of 12 TB drives strands 8 TB on each larger drive. The calculator above shows that stranded figure explicitly because it is the number vendor tools are least likely to display.
How we choose: we compare published manufacturer specifications, drive datasheets, published reliability statistics and verified owner reviews. We do not test hardware in person, and we are not tied to any NAS vendor. Capacity and power figures here are researched guidance, not a warranty. RAID protects against drive failure, not against deletion, ransomware, fire or theft, so keep verified backups regardless of what any calculator tells you.
Working out your own cost per usable terabyte? The Home Server Build Planner is the paid version of these pages: 8 printable worksheets you fill in with your own numbers, plus the full PDF, $29.