RAID capacity and redundancy calculator
Compare usable capacity, efficiency, redundancy overhead and drive-failure tolerance for common RAID layouts.
What RAID actually does
RAID — Redundant Array of Independent Disks — combines multiple storage devices using techniques such as striping, mirroring and parity. Depending on the RAID level, the goal can be higher throughput, fault tolerance, capacity efficiency or a compromise among them. RAID level alone does not describe every performance characteristic: controller, media, cache, workload, stripe geometry and implementation matter.
RAID 0: maximum capacity, no redundancy
RAID 0 stripes data across multiple drives and uses the full combined capacity. There is no parity or mirror copy. This can provide high sequential and parallel I/O potential, but any member-drive failure can make the entire array unusable. RAID 0 should therefore never be interpreted as data protection.
RAID 1: mirroring and simple recovery
RAID 1 stores complete copies of the same logical data on independent drives. With two equal drives, usable capacity is the capacity of one drive. Mirroring avoids parity reconstruction and is operationally straightforward, but capacity efficiency is lower. This calculator also allows multi-way mirroring for educational capacity modeling; actual controller support varies.
RAID 5: distributed parity and one-drive fault tolerance
RAID 5 stripes data and distributed parity across the member devices. Its usable-capacity model is approximately (N − 1) × the smallest-drive capacity. It can survive one member-device failure without data loss, but the array operates degraded until reconstruction is complete. Write workloads also incur parity-related work.
RAID 6: dual parity and two-drive fault tolerance
RAID 6 extends parity protection so stored data can survive any two member-device failures. Its common usable-capacity model is (N − 2) × the smallest-drive capacity. The extra protection costs another drive-equivalent of capacity and additional parity work compared with RAID 5.
RAID 10: stripe over mirrored pairs
RAID 10, also called RAID 1+0, stripes data across mirrored sets. In the conventional two-way mirror model used here, usable capacity is 50% of raw capacity. It guarantees survival of one drive failure and can survive multiple simultaneous failures when they occur in different mirror pairs. If both members of the same mirror pair fail, the array cannot reconstruct that portion of the data.
RAID 10 and RAID 0+1 are not the same layout
The names are sometimes confused. RAID 10 (1+0) creates a stripe across mirrored devices or mirror sets. RAID 0+1 mirrors striped sets. They can have similar raw capacity, but their failure behavior and reconstruction topology differ. When a product says “RAID 10”, verify the controller documentation rather than assuming every nested RAID implementation behaves identically.
Mixed drive sizes: why the smallest drive matters
The formulas in this calculator assume equal-size members. In many conventional RAID implementations, when drives of different capacities are combined, only an amount comparable to the smallest member is usable from each drive for the RAID set. Vendor-specific technologies may use mixed capacities differently, so always validate the storage platform documentation.
Why a “4 TB” drive may appear smaller in the operating system
Storage manufacturers normally advertise decimal capacity: 1 TB = 1,000,000,000,000 bytes. Binary units use powers of two: 1 TiB = 1,099,511,627,776 bytes. A 4 TB drive therefore contains the advertised number of bytes but corresponds to about 3.64 TiB. Filesystem, metadata, reserved space and storage-platform overhead can reduce the space visible to applications further.
A degraded RAID is not back to normal until rebuild finishes
When a protected array loses a drive, redundancy is reduced until the missing data has been reconstructed onto a replacement or spare. Rebuild can place substantial I/O load on the remaining members and can take significant time on large arrays. Actual duration depends on drive technology, capacity, workload, controller behavior, rebuild priority and implementation.
A hot spare does not increase normal usable capacity
A dedicated hot spare is normally kept ready to replace a failed member automatically or with minimal intervention. Because it is not part of the active data layout during normal operation, it should not be counted as usable RAID capacity. Distributed-spare and vendor-specific designs can work differently.
Redundancy and backup solve different problems
RAID can improve service continuity after certain device failures, but every mirror or parity member still represents the same logical data set. Accidental deletion, ransomware, application corruption, filesystem damage, administrator error, theft, fire or a catastrophic storage-system failure can affect the array as a whole. Maintain independent, recoverable and tested backups.
RAID performance cannot be predicted from capacity math alone
Simple statements such as “RAID 5 is fast” or “RAID 10 is always faster” omit important variables. Sequential versus random access, read/write ratio, queue depth, HDD versus SSD/NVMe, controller cache, stripe size, filesystem alignment and parity implementation can change results substantially. Treat the performance labels in this calculator as qualitative orientation, not benchmark predictions.
RAID terminology varies across storage platforms
Hardware RAID controllers, Linux software RAID, NAS appliances, hyperconverged systems and filesystems such as ZFS can expose different layouts and protection models. RAIDZ, distributed erasure coding, declustered RAID and vendor-specific hybrid arrays should not automatically be calculated with the classic RAID formulas on this page. Use the native documentation and capacity model for those technologies.