RAID combines multiple drives into a logical volume — but which configuration matches your requirements? The choice between RAID 0, RAID 1, RAID 5, RAID 6 and RAID 10 determines usable capacity, performance, rebuild duration and failure tolerance. This article explains how each RAID level works, which deployment scenarios they serve, and how to weigh capacity cost against rebuild risk.
What Does RAID Do?
RAID (Redundant Array of Independent Disks) distributes data across multiple drives to improve performance, provide redundancy, or both. The RAID level determines whether data is striped for throughput, mirrored for immediate redundancy, or protected by parity for capacity-efficient fault tolerance.
A RAID controller manages the array: it writes data according to the chosen scheme, rebuilds the array when a drive fails, and presents the array to the operating system as a single logical disk. Hardware RAID controllers offload parity calculations from the CPU and include battery-backed or flash-backed write cache to protect in-flight writes during a power loss. RAID is not a backup. It protects against drive failure, not against data corruption, accidental deletion, malware or fire. A RAID array and an independent backup together form a resilient storage strategy.
RAID 0 — Striping Without Redundancy
RAID 0 stripes data evenly across all drives. Every block is split into segments and written in parallel, so read and write throughput scales with the number of drives. Usable capacity equals the sum of all drive capacities.
The penalty: RAID 0 offers no redundancy. If one drive fails, the entire array is lost. RAID 0 is therefore unsuitable for production data.
When RAID 0 Makes Sense
RAID 0 is appropriate only for workloads where performance matters and data loss is acceptable — for example scratch space during rendering, temporary datasets that can be regenerated, or staging environments where the canonical copy lives elsewhere. Because it offers no protection, RAID 0 should never hold business-critical data, databases, or anything that cannot be rebuilt from source within the acceptable downtime window.
Minimum configuration: two drives. Common drive counts: 2, 4, 6 or 8, depending on the required throughput.
RAID 1 — Mirroring for Immediate Redundancy
RAID 1 writes every block identically to two drives. Each drive holds a complete copy of the data, so usable capacity equals the size of a single drive. If one drive fails, the other continues to serve all requests without interruption.
Rebuild is straightforward: the controller copies the surviving drive onto the replacement. Because there is no parity calculation, the rebuild proceeds at the sustained write speed of the new drive.
When RAID 1 Is the Right Choice
RAID 1 suits workloads that demand high read performance, low rebuild risk, and simple recovery. It is common for boot volumes, hypervisor datastores, database logs and any dataset where simplicity and fast rebuild outweigh capacity cost. RAID 1 doubles the cost per usable gigabyte, so it is less economical than parity-based levels for large datasets. RAID 1 can be extended to three or more mirrors (RAID 1E or vendor-specific implementations), but two-way mirroring remains the standard configuration.
Minimum configuration: two drives. RAID 1 scales poorly for large capacity; beyond two drives, RAID 10 or parity-based levels are more economical.
RAID 5 — Parity Across Three or More Drives
RAID 5 stripes data and parity evenly across all drives. For every stripe, one segment is parity; the rest is data. If one drive fails, the missing blocks are reconstructed in real time from the surviving data and parity. Usable capacity equals (n – 1) × drive size, where n is the number of drives. A four-drive RAID 5 array yields three drives’ worth of space; a six-drive array yields five. RAID 5 tolerates the loss of one drive.
RAID 5 Rebuild Risk
Rebuild duration scales with array size. During rebuild, the controller reads every block from every surviving drive and recalculates the missing data. On large arrays, rebuild can take many hours or days. While the rebuild runs, the array operates in degraded mode: if a second drive fails before rebuild completes, the array is lost.
Unrecoverable read errors (URE) compound the risk. Enterprise drives typically specify a URE rate of 10^-15, which statistically corresponds to approximately one unrecoverable read error per 125 TB of data read. On a rebuild of a six-drive array with 10 TB drives, the controller reads 50 TB of data — statistically, the probability of encountering at least one URE during rebuild is significant. A URE during RAID 5 rebuild causes data loss. RAID 5 is widely regarded as unsuitable for arrays larger than a few terabytes or for drives above 4 TB. RAID 6 provides better protection for the same workload.
When RAID 5 Is Still Appropriate
RAID 5 remains viable for small arrays (four to six drives, each under 2 TB), short rebuild windows, and workloads where the data changes infrequently enough that a second failure during rebuild is unlikely. It is more capacity-efficient than RAID 10 and simpler than RAID 6, but those advantages diminish as drive sizes grow.
Minimum configuration: three drives. Typical deployment: four to six drives, each 1–2 TB, for file shares, archive storage or environments where capacity efficiency outweighs rebuild risk.
RAID 6 — Dual Parity for Large Arrays
RAID 6 extends RAID 5 by calculating two independent parity blocks per stripe. Usable capacity equals (n – 2) × drive size. The array tolerates the simultaneous loss of any two drives, or the loss of one drive plus a URE during rebuild.
Why RAID 6 Became Necessary
As drive capacities grew into the terabyte range, the rebuild duration and URE probability made RAID 5 increasingly risky for large arrays. RAID 6 provides a second layer of protection precisely where RAID 5’s single-parity scheme fails: during rebuild. Even if a URE occurs, or a second drive fails while the first is being rebuilt, the array survives. The write penalty is higher than RAID 5 because every write updates two parity blocks. Read performance is comparable. Rebuild duration is similar to RAID 5, but the array remains protected throughout.
When RAID 6 Is the Right Choice
RAID 6 is generally the preferred parity-based level for arrays using drives above 4 TB, for large drive counts (eight or more), and for any dataset where a second failure during rebuild would cause unacceptable data loss. It is common in file servers, bulk storage, backup targets and archive systems where capacity efficiency and rebuild safety both matter. A RAID 6 array with six 10 TB drives yields 40 TB usable — the same raw capacity as a four-drive RAID 10 array, but with lower cost per gigabyte and tolerance for two failures instead of one.
Minimum configuration: four drives. Typical deployment: six to twelve drives, each 4–18 TB, for any workload where RAID 5 rebuild risk is unacceptable.
RAID 10 — Mirroring Plus Striping
RAID 10 combines RAID 1 and RAID 0: drives are paired into mirrors, and the mirrors are striped. Usable capacity equals half the raw capacity. The array tolerates the loss of one drive per mirror pair. If both drives in any pair fail, the array is lost. Read and write performance scale well with drive count. Because rebuild only requires copying a single drive, rebuild is faster than parity-based levels and imposes no parity calculation overhead on the controller or the surviving drives.
When RAID 10 Outperforms RAID 6
RAID 10 suits workloads that demand high random I/O, consistent low latency, and fast rebuild: databases, virtualisation datastores, transactional systems and any application where the performance penalty of parity calculation is unacceptable. The capacity cost is higher than RAID 6, but the performance and rebuild characteristics are predictable. For a given number of drives, RAID 10 provides better IOPS than any parity-based level. The trade-off is usable capacity: a twelve-drive RAID 10 array yields six drives’ worth of space; a RAID 6 array yields ten.
Minimum configuration: four drives (two mirror pairs). Typical deployment: four to twelve drives, often enterprise NVMe SSDs or high-RPM SAS drives, for databases, hypervisors or high-throughput applications.
Comparing Usable Capacity and Rebuild Behaviour
| RAID Level | Drives | Usable Capacity | Failures Tolerated | Rebuild Stress | Typical Use Case |
|---|---|---|---|---|---|
| RAID 0 | 2+ | 100% | None | n/a | Scratch space, staging, non-critical data |
| RAID 1 | 2 | 50% | 1 | Low (copy only) | Boot volumes, logs, small high-value datasets |
| RAID 5 | 3+ | (n-1)/n | 1 | High (full stripe read) | Small arrays, infrequent writes, legacy deployments |
| RAID 6 | 4+ | (n-2)/n | 2 | High (full stripe read) | Large arrays, archival, bulk storage |
| RAID 10 | 4+ | 50% | 1 per pair | Low (copy only) | Databases, virtualisation, transactional workloads |
Rebuild stress refers to the controller load and the quantity of data read from surviving drives during reconstruction. Parity-based levels must read every block on every surviving drive; mirrored levels copy a single drive.
How RAID Level Affects Controller and Drive Selection
The RAID level determines which RAID controller capabilities matter most. RAID 5 and RAID 6 require hardware with sufficient cache and processing power to handle parity calculations without bottlenecking write performance. Controllers used in RAID 6 arrays benefit from battery-backed or supercapacitor-backed write cache to protect parity updates during a power loss. RAID 10 places less demand on the controller’s computational resources but requires more drive bays to achieve the same usable capacity as a parity-based array. Drive choice also shifts: RAID 10 deployments often use faster, smaller-capacity drives to maximise IOPS, while RAID 6 deployments prioritise capacity and endurance. Hot spare configuration is independent of RAID level but affects rebuild behaviour. A dedicated hot spare allows rebuild to begin immediately upon failure; without one, rebuild starts only after a replacement drive is installed and detected.
Size RAID Arrays According to the Workload
Choosing a RAID level requires balancing capacity efficiency, performance, rebuild risk and cost. For any given workload, establish:
- Capacity requirement — how much usable space is needed, and how does that translate into raw drive count per RAID level?
- Performance requirement — does the workload demand high sequential throughput, high random IOPS, or low latency? Parity-based levels reduce write performance; RAID 10 scales IOPS linearly.
- Failure tolerance — can the system tolerate a rebuild window of hours or days? Is a second failure during rebuild acceptable?
- Drive size and count — larger drives extend rebuild duration and increase URE probability. Beyond 4 TB per drive, RAID 6 or RAID 10 are safer than RAID 5.
- Data criticality — can the data be regenerated, or does its loss halt the business? RAID does not replace backups, but critical data demands a fault-tolerant RAID level.
The highest number is not the best choice. RAID 6 is safer than RAID 5, but RAID 10 may be faster and simpler to rebuild. RAID 0 maximises capacity and performance, but loses everything on a single failure. The right decision follows from the workload, not from a hierarchy of RAID levels.
When planning a RAID system, the RAID level, controller, drives, and server platform must be coordinated with one another. On server-hardware.com, businesses and system integrators can find suitable servers and storage components, such as storage devices. Are you unsure which RAID configuration is right for your requirements? Our team will assist you in selecting compatible components and the right server configuration!
Select the RAID Level the Workload Demands
RAID levels trade capacity, performance and rebuild safety against one another. RAID 0 maximises throughput but offers no protection. RAID 1 mirrors data at 50% capacity efficiency and rebuilds quickly. RAID 5 saves capacity but exposes large arrays to rebuild risk. RAID 6 tolerates two failures and is the safe choice for multi-terabyte arrays. RAID 10 provides the best random I/O performance and the simplest rebuild, at the cost of usable capacity. The right choice is not the highest number. It is the level that fits the workload’s performance profile, failure tolerance, capacity budget and rebuild risk. For databases and virtualisation, RAID 10 typically outperforms parity-based levels. For archival storage and large file systems, RAID 6 provides the best balance of capacity and safety. For boot volumes and small critical datasets, RAID 1 remains the simplest and most reliable option. RAID does not replace backups. It protects against hardware failure; backups protect against everything else. A production storage strategy requires both.
Are you planning a RAID system? Find suitable servers, RAID controllers, and storage components at server-hardware.com or get individual advice!
Online Marketing Manager
