RAID Calculator

Free RAID calculator for capacity, efficiency, and fault tolerance

Here is the short version: usable space is always less than raw except on RAID 0, and RAID 0 is not protection. This RAID calculator estimates usable capacity, efficiency, fault tolerance, and a first-pass IOPS range for RAID 0, 1, 1E, 5, 6, 10, 50, and 60 before you buy disks.

Use it when you are quoting a NAS, a hypervisor datastore, or a backup repository and you need a number you can defend. It assumes equal-sized disks. Vendor hybrids (Synology SHR, unRAID, Windows Storage Spaces) can differ. Microsoft documents the software-defined cousin in Storage Spaces overview.

RAID formulas used (equal-sized disks)

Let n = data disks (excluding hot spares) and S = size of each disk.

Hot spares count toward the buy list. They do not add usable capacity until they replace a failed disk.

Which RAID level for which job

Pick the workload first, then the level.

• SQL, VMs, and busy file servers: RAID 10. Rebuilds are a mirror copy, not a parity reconstruction across the whole pack. Writes stay livable.
• Bulk backup or archive: RAID 6 or RAID 60. Capacity efficiency matters. Dual parity survives a second failure during a long rebuild.
• Small two-disk appliance: RAID 1. Simple. Capacity equals one disk.
• Scratch or transcode cache: RAID 0 only if you accept total loss. Never put the only copy of a client matter there.
• Large disk counts: RAID 50 or 60 isolates rebuilds to a group. Only bother when you have the disk count and a reason.

RAID 5 is still common. On 16 TB and 20 TB spinning disks the rebuild window is long enough that a second failure or an unrecoverable read is a real planning risk. Prefer RAID 6 or 10 for production data on large HDDs.

Performance notes (approximate)

Read and write multipliers in the results panel are educational:

• RAID 0 and RAID 10 scale reads well. RAID 10 writes scale with the number of mirror pairs.
• RAID 5 and 6 pay a parity write penalty. Classic teaching models use about 4× and 6× operation cost. That is why random-write databases feel slow on parity arrays.
• Real IOPS depend on disk type, queue depth, controller cache, and whether the set is SSD or HDD. Enter your disk IOPS to scale the estimate. Do not treat the IOPS line as a vendor quote.

SSD arrays still need a RAID plan. They rebuild faster, which lowers (does not remove) the second-failure risk. They also wear. Watch spare capacity and endurance ratings.

Rebuild risk on large disks

Parity RAID rebuilds read almost the entire remaining array. As multi-terabyte disks became common, long rebuild windows increased the chance of a second failure or an unrecoverable read error (URE). The rebuild-risk note uses your URE exponent as a teaching signal, not a guarantee from any drive vendor.

A hot spare shortens the wait for a human to swap a disk. It does not make RAID 5 safe on huge HDDs. Dual parity or striped mirrors do more for that problem.

RAID is not backup

An array can survive a disk failure and still lose everything to ransomware, accidental deletion, fire, or a failed controller. CISA’s data backup guidance is the same lesson in public-sector language: copies that cannot be changed, stored somewhere else, and tested.

Pair RAID with:

• Versioned, immutable or offline backups
• Restore tests on a schedule
• Monitoring for a degraded array (not “we will notice”)
• Documented RTO and RPO

See backup and disaster recovery and managed infrastructure. Camera systems should also use the surveillance storage calculator. Size the hosts that sit on the array with the VM calculator.

Mixing disk sizes and other traps

Traditional RAID is limited by the smallest disk. Mixing a 4 TB and an 8 TB usually wastes 4 TB per larger disk unless the vendor offers a hybrid mode. Do not mix SSD and HDD in one RAID set unless the vendor designed that tier.

Other traps:

• Two RAID 0 disks “because we have backups” and then skipping the backup job
• Forgetting the hot spare in the purchase quote
• Building RAID 5 on a USB dock and calling it a SAN
• Ignoring write-cache battery or equivalent on a hardware controller

Worked examples

Six 4 TB disks in RAID 6, one hot spare

• Data disks n = 6, S = 4 TB
• Usable = (6 − 2) × 4 = 16 TB
• Raw data pool = 24 TB, efficiency about 67%
• Total drives to buy = 7 (6 + spare)

Eight 1.92 TB SSDs in RAID 10

• n = 8, even, S = 1.92 TB
• Usable = (8 / 2) × 1.92 = 7.68 TB
• One disk can fail in each mirror pair without data loss
• Better write behavior for VMs than the same eight disks in RAID 5

Buying the disks without wasting a purchase order

Write three numbers on the quote: usable TB you need in 18 months, RAID level, and spare count. Then work backward. If you need 12 TB usable for files plus growth, RAID 6 on 4 TB disks is four data disks of usable after parity… wait: (n-2)×4 = 12 means n = 5, so five 4 TB data disks plus a spare is six drives. Run those numbers in the calculator before you accept a vendor bundle.

Prefer one disk model and firmware family. Mixing leftover drives from a desk drawer is how arrays degrade on a Thursday. For hypervisor datastores, size usable space after snapshots. For backup repositories, remember the backup software’s own overhead.

Watch the array after go-live. A silent degraded RAID 6 is still one failure from a bad week. Alerts should page a person, not sit in an unread mailbox.

When to call Secure Techies

If you are designing SAN or NAS for VMs, SQL, file shares, or backup repositories, we can size the array, monitoring, and recovery plan together. Contact us for a free assessment.