RAID Usable Capacity Calculator
Turn a shelf of equal-size drives into a transparent capacity plan. Compare RAID 0, RAID 1, RAID 5, RAID 6, and RAID 10 while keeping hot spares, redundancy, decimal-versus-binary units, and controller or filesystem reserve visible.
Configure one uniform-drive array
All active drives are modeled at the entered per-drive capacity. If real drives differ, enter the smallest usable member size.
One installed drive is reserved as a hot spare. The seven active members use one drive-equivalent for distributed parity, leaving six data-equivalent drives before the 5% post-RAID reserve.
Capacity equations by RAID level
This calculator assumes a single array made from drives with one common usable member size. The active count is installed drives minus dedicated hot spares. RAID protection then determines how many active drive-equivalents can hold user data. A post-RAID percentage is finally removed to represent controller metadata, filesystem formatting, snapshots, thin-provisioning reserve, or intentional operational free space.
RAID data capacity = data-equivalent drives × smallest drive size
Presented estimate = RAID data capacity × (1 − reserve percentage)
TiB = decimal TB × 10¹² ÷ 2⁴⁰
| Level | Data-equivalent drives from N active members | Minimum active members | Simplified failure tolerance |
|---|---|---|---|
| RAID 0 | N | 2 in this calculator | None; one member failure can make the array unavailable. |
| RAID 1 | 1 | 2 | All but one mirror member, assuming the remaining copy is readable. |
| RAID 5 | N − 1 | 3 | One active member. |
| RAID 6 | N − 2 | 4 | Two active members. |
| RAID 10 | N ÷ 2 | 4 and even | At least one member per mirror pair; multiple failures survive only when they are in different pairs. |
These equations estimate capacity, not performance or probability of data loss. Controller implementations can use different layouts, chunk sizes, distributed spare schemes, parity groups, pools, declustered RAID, or erasure-code notation. Confirm the exact scheme with the storage platform documentation before purchasing equipment.
TB is not TiB
Drive manufacturers generally advertise decimal capacity: one terabyte equals 1,000,000,000,000 bytes. Many operating systems and storage tools report a binary quantity based on 1,099,511,627,776 bytes, properly called a tebibyte or TiB, even when the interface labels it “TB.” No bytes vanish during that conversion. The same byte count is divided by a larger unit, so the numeric TiB result is smaller.
In the default example, 68.40 decimal TB is about 62.21 TiB. Treating those displays as two separate losses would double-count the unit difference. The capacity waterfall performs RAID and reserve arithmetic in decimal TB, then provides TiB as an alternate display of the final bytes.
Real per-drive usable capacity can also differ slightly because of sector formatting, overprovisioning, zoned-drive rules, controller qualification, or a vendor’s definition of member capacity. When mixing nominally similar products, use the smallest controller-recognized member size. Capacity beyond the smallest member is commonly unusable within a conventional uniform RAID group.
Hot spare is readiness, not parity
A dedicated hot spare does not normally hold active array data before a failure, so this model removes it before RAID capacity is calculated. When a member fails, the controller can begin rebuilding onto the spare without waiting for a technician to install a replacement. That shortens exposure time but does not change the RAID level’s mathematical concurrent-failure tolerance.
For example, RAID 5 still has single-parity protection. A hot spare can receive reconstructed data after the first failure, but it is not a second parity member. If another active member fails before reconstruction finishes, the array may be lost. RAID 6 carries two parity-equivalent members and can tolerate two active-member failures under its model, though additional latent media errors and implementation issues still matter.
Some systems use distributed spare capacity rather than whole dedicated spare drives. In that case, enter an equivalent spare count only for a rough estimate or model the vendor’s documented usable capacity directly. Do not assume the rebuild speed, fault domain, or resiliency of distributed sparing is identical to one dedicated drive.
Worked eight-drive RAID 5 example
Suppose a U.S. small-business storage server has eight advertised 12 TB drives. One is assigned as a dedicated hot spare, leaving seven active RAID members. RAID 5 uses one active drive-equivalent for distributed parity, so six drive-equivalents remain for data. Six multiplied by 12 TB produces 72.00 TB before filesystem and operational reserve.
Entering 5% post-RAID reserve removes 3.60 TB, leaving an estimated 68.40 TB presented. That byte count is 68,400,000,000,000 bytes, or approximately 62.21 TiB. Relative to all eight purchased drives, estimated presented efficiency is 71.25%. The capacity not presented consists of 12 TB in the spare, 12 TB of parity-equivalent protection, and 3.60 TB of post-RAID reserve.
The result does not promise that a particular NAS or SAN will show exactly 68.40 TB. Pool metadata, filesystem structure, checksums, snapshots, deduplication tables, block size, reserved free space, thin provisioning, and vendor display conventions can change the number. Use the calculator as a planning ledger, then compare it with the vendor’s sizing tool and configuration preview.
Capacity is only one design axis
RAID level affects usable capacity, write behavior, rebuild work, fault tolerance, and failure exposure. Workload matters: random writes, large sequential streams, virtual machines, databases, backups, and surveillance archives have different latency and throughput patterns. Media type, queue depth, controller cache policy, stripe geometry, network speed, and filesystem can dominate observed performance.
Large drives can take substantial time to rebuild. During degraded operation, surviving members may experience heavy read activity and reduced redundancy. Reliability planning should consider annualized failure behavior, correlated failures, error handling, scrubbing, patrol reads, environmental conditions, firmware, replacement logistics, and the size of each protection group. Capacity efficiency alone is not a safe reason to choose single parity.
RAID is also not a backup. Mirroring and parity can maintain service after certain device failures, but they reproduce or preserve logical damage such as accidental deletion, ransomware encryption, application corruption, or an administrator error. Maintain independent, tested backups with appropriate offline or immutable copies, retention, access control, and recovery drills. SNIA’s mirroring guidance explicitly distinguishes hardware-failure resilience from protection against compromised, deleted, or overwritten data.
Planning checklist before ordering drives
- Confirm the controller’s supported RAID layouts, maximum members, drive qualification list, sector format, and maximum recognized capacity.
- Use the smallest actual member capacity when models or firmware revisions differ.
- Separate hot spares from active members and confirm whether the platform uses dedicated, global, or distributed sparing.
- Estimate filesystem, metadata, snapshot, replication, and free-space policies instead of assuming raw RAID capacity is fully allocatable.
- Size for growth and retention while maintaining the free space required for stable performance and recovery operations.
- Validate workload performance and rebuild behavior, not just steady-state sequential throughput.
- Document fault domains across enclosures, power supplies, controllers, cables, and sites; multiple drives are not useful protection when a shared component removes them together.
- Design and test a separate backup and restore process.
RAID usable capacity FAQs
Why does an eight-drive RAID 5 with one spare have only six data drives?
The spare is not an active data member, leaving seven active drives. RAID 5 uses one drive-equivalent of those active members for distributed parity, so capacity equals six times the smallest drive size before post-RAID overhead. Parity blocks are distributed; there is not necessarily one physical “parity drive.”
Does RAID 1 with four drives provide two drives of capacity?
Not in this calculator’s RAID 1 model. It treats all active members as one mirror set containing identical copies, so usable RAID capacity is one smallest drive. Two independent mirrored pairs would be two RAID 1 groups, or they could participate in RAID 10 if striped across pairs.
Can RAID 10 survive two drive failures?
Sometimes, but placement matters. It can survive multiple failures when no mirror pair loses all its members. Two failed drives in different pairs can be survivable; two failures in the same two-member mirror pair are not. The result therefore states at least one member per pair rather than promising any two failures.
Why does my operating system show fewer terabytes?
Part may be the TB-versus-TiB display difference, and part may be real overhead or reserve. Controller metadata, pool structures, filesystems, snapshots, checksums, and free-space policy can reduce allocatable capacity. Compare byte counts and documented boundaries before concluding capacity is missing.
Should I count a global hot spare once for every array?
No. A global spare may be shared across several eligible arrays. Allocating its entire capacity as a loss in each array would double-count it. Model the full storage system or apportion the spare deliberately, then confirm which arrays, media types, and enclosures the controller permits it to protect.
Is RAID 5 safe for large drives?
A capacity calculator cannot answer that. Safety depends on workload, rebuild duration, media error behavior, controller implementation, scrubbing, failure correlation, backup quality, and acceptable risk. Evaluate vendor guidance and recovery requirements; higher capacity efficiency is only one factor in choosing a protection scheme.
References
These vendor-neutral SNIA resources support the RAID, mirroring, parity, and resiliency concepts used in this planning calculator. The selected storage platform’s current documentation controls an actual configuration.