Comparison · RAID layouts
Usable space, failures survived, speed and rebuild load, worked out on the drive counts a storage server really has.
Quick answer
RAID 6 or RAID 10: which is better for a rented storage server?
For a rented storage server, RAID 6 is better for capacity, RAID 10 for random writes. On ten 10 TB drives, RAID 6 keeps 80 TB and survives any two failures; RAID 10 keeps 50 TB and survives one per pair. Basalthold Hold 100 has those drives for $377.05 per 30 days (2 November 2026), crypto only, no KYC.
Usable capacity in decimal TB, before filesystem overhead. RAIDZ2 in ZFS holds about the same as RAID 6.
| Drives | RAID 6 or RAIDZ2 usable | RAID 10 usable | RAID 6 survives | RAID 10 survives |
|---|---|---|---|---|
| 2 × 12 TB | Not possible: RAID 6 needs 4 drives, RAIDZ2 needs 3 | 12 TB as a two-way mirror | — | 1 drive |
| 4 × 10 TB | 20 TB | 20 TB | Any 2 drives | 1 per pair; 2 if they sit in different pairs |
| 10 × 10 TB | 80 TB | 50 TB | Any 2 drives | 1 per pair; up to 5 if each hits a different pair |
| 10 × 16 TB | 128 TB | 80 TB | Any 2 drives | 1 per pair; up to 5 if each hits a different pair |
| 10 × 20 TB | 160 TB | 100 TB | Any 2 drives | 1 per pair; up to 5 if each hits a different pair |
RAID 6 = (N − 2) × drive size; RAID 10 with two copies = N ÷ 2 × drive size. Drives in the holds: 2 × 12 TB HDD, 4 × 10 TB HDD, 10 × 10 TB HDD, 10 × 16 TB HDD.
RAID 6 gives up two drives to parity, whatever the array size. RAID 10 gives up half the drives to copies. The Linux RAID wiki puts it simply: (N − 2) × S for RAID 6, where S is the smallest drive.
On four drives the two layouts tie. On ten drives, RAID 6 keeps 80% of the raw space and RAID 10 keeps 50%.
On ten drives, RAID 10 gives up three drives' worth of space to buy write speed.
Price the usable terabyte, not the raw one. Divide the rent by the usable space: $377.05 ÷ 80 for RAID 6 on Hold 100, or ÷ 50 for RAID 10. A whole 100 TB machine is described on the 100 TB storage server page.
RAID 6, by design. The Linux md documentation says RAID 6 can handle the loss of any two devices without data loss. RAID 10 survives one loss in each mirror pair, but losing both drives of one pair ends the array.
Take ten drives in five pairs. After the first failure, one of the nine remaining drives is fatal if it fails next: its partner. In RAID 6, any second failure is survivable, and only a third is fatal.
The rebuild read matters too. Seagate rates the Exos X20 at one unreadable sector per 10^15 bits read, at most. Rebuilding RAID 10 reads one 10 TB partner, 8 × 10^13 bits. Rebuilding RAID 6 reads nine drives, 7.2 × 10^14 bits. RAID 6 reads more, but it still has a second parity to repair a bad sector. Rebuild durations are worked out on the RAID rebuild time page.
RAID 10 for small random writes, RAID 6 for capacity per drive. A RAID 10 write goes to the two drives of a pair. A small RAID 6 write must update the data block and two parity blocks.
The md manual adds a warning: RAID 6 is "very slow in dual disk failure mode". With one drive gone, it is comparable to RAID 5.
For most storage-server jobs the network decides. Backups, archives and media arrive as long sequential streams through a 1 Gbit/s port, at 125 MB/s at most. One modern hard drive reads faster than that on its outer tracks, so either layout keeps up. Random work is different; see HDD or SSD.
RAIDZ2 is the ZFS version of RAID 6. A raidz group of N disks with P parity disks holds about (N − P) × X and survives P failures. OpenZFS adds that raidz does not suffer from the RAID 5/6 "write hole", where data and parity disagree after a power loss.
OpenZFS recommends 3 to 9 disks per raidz group for performance. Ten drives can form one 10-wide RAIDZ2 for capacity. Two 5-wide RAIDZ2 groups give more random speed but keep 60 TB of 100 TB raw.
The ZFS equivalent of RAID 10 is a pool of mirrors. Which engine to use is on ZFS or hardware RAID.
Pick by the workload and the drive count.
You build the array yourself, or ask for it. The operating system sits on its own SSD (two mirrored SSDs on the larger holds); the data drives arrive blank and unpartitioned, so you can build ZFS, an mdadm array or a set of independent disks. If you want the team to prepare a RAID 6 or RAID 10 array before delivery, say so to support before the server is handed over.
Of 23 storage-server sellers Basalthold checked on 2 November 2026, only 4 stated a hardware RAID controller on a storage plan. Most arrays in this market are software arrays.
RAID keeps a server running through a failed drive. It does not protect against deletion or ransomware: A storage server is not a backup of itself: we keep no copy of your data anywhere else, so a second copy in another place stays your job.
Hold 24 suits a mirror; Hold 40 and up take RAID 6, RAID 10 or RAIDZ2. Longer terms are on the price list.
24 TB raw · 6 cores
$110.99
$99.89 / month
$299.67 for 3 months
$88.79 / month
$532.75 for 6 months
$77.69 / month
$932.31 for 1 year
3 months $299.67 (−10%)6 months $532.75 (−20%)1 year $932.31 (−30%)
40 TB raw · 6 cores
$201.09
$180.98 / month
$542.94 for 3 months
$160.87 / month
$965.23 for 6 months
$140.76 / month
$1,689.15 for 1 year
3 months $542.94 (−10%)6 months $965.23 (−20%)1 year $1,689.15 (−30%)
100 TB raw · 10 cores
$377.05
$339.34 / month
$1,018.03 for 3 months
$301.64 / month
$1,809.84 for 6 months
$263.94 / month
$3,167.22 for 1 year
3 months $1,018.03 (−10%)6 months $1,809.84 (−20%)1 year $3,167.22 (−30%)
Most TB per dollar
160 TB raw · 16 cores
$561.39
$505.25 / month
$1,515.75 for 3 months
$449.11 / month
$2,694.67 for 6 months
$392.97 / month
$4,715.67 for 1 year
3 months $1,515.75 (−10%)6 months $2,694.67 (−20%)1 year $4,715.67 (−30%)
In every plan: Blank data drives · Encryption keys stay yours · Dedicated IPv4 · No logs · No KYC · No setup fee
Yes, if you ask before delivery. The operating system sits on its own SSD (two mirrored SSDs on the larger holds); the data drives arrive blank and unpartitioned, so you can build ZFS, an mdadm array or a set of independent disks. If you want the team to prepare a RAID 6 or RAID 10 array before delivery, say so to support before the server is handed over.
It survives any two failures, which is the point with large drives. The weak spot is the rebuild: a 20 TB drive takes about 19.5 hours at its top speed, often far longer under load. Keep a second copy elsewhere.
When a drive fails, open a ticket: the team has it replaced at no cost, and rebuilding the array on the new drive is up to your setup. With mdadm you add the new drive to the array; with ZFS you run zpool replace.
Divide the rent by the usable space. Hold 160 costs $561.39 per 30 days with 10 × 16 TB HDD; RAID 6 keeps all of them but two. The raw figure is $3.50/TB, so the usable figure is higher.
Yes. Basalthold takes BTC, USDT, USDC, ETH, SOL, LTC, from $25. No KYC: no ID, no selfie, no documents, no card and no phone number, at any point. Setup takes 72 hours at the most after your payment is confirmed.
Build your own RAID 6 on blank drives
Hold 100: Intel Xeon Silver 4210R (10 cores / 20 threads), 128 GB DDR4 ECC, 10 × 10 TB hard drives (100 TB raw) plus 2 × 480 GB SSD for the system, mirrored, 1 Gbit/s port with 120 TB of traffic per 30 days, 1 IPv4 and an IPv6 /64, for $377.05/30 days. Crypto, no KYC.
24 TB from $110.99/30 days · no KYC
Choose a hold