The same set of drives can behave very differently depending on how a controller groups them. One layout maximizes speed but offers zero protection, another survives two simultaneous drive failures, and another doubles your drive cost in exchange for predictable performance even while degraded. Picking the right RAID level is one of the highest-impact storage decisions you make when building or upgrading a server.
This guide explains the five RAID levels that cover the vast majority of real deployments — RAID 0, 1, 5, 6 and 10 — in practical terms: usable capacity, fault tolerance, write penalties and rebuild behavior, plus clear recommendations for which workloads each level fits.
What RAID Does (and What It Does Not)
RAID (Redundant Array of Independent Disks) combines multiple physical drives into one logical volume to improve performance, availability, or both. Two mechanisms underpin every level: striping, which splits data across drives so several spindles or flash channels work in parallel, and redundancy, which stores mirror copies or parity so the array keeps running when a drive dies.
One thing RAID is not: a backup. RAID protects against drive hardware failure and nothing else. Accidental deletion, ransomware, file-system corruption and controller failure all pass straight through to a "protected" array. Every serious deployment pairs RAID with an independent backup — for many businesses that still means LTO tape, which provides offline, air-gapped copies at the lowest cost per terabyte.
RAID Levels at a Glance
| Level | Minimum drives | Usable capacity | Failures tolerated | Best for |
|---|---|---|---|---|
| RAID 0 | 2 | 100% | 0 | Scratch and temp data only |
| RAID 1 | 2 | 50% | 1 | OS and boot volumes |
| RAID 5 | 3 | N−1 drives | 1 | Read-heavy arrays of smaller drives |
| RAID 6 | 4 | N−2 drives | 2 | Large-capacity storage and backup targets |
| RAID 10 | 4 | 50% | 1 per mirror pair | Databases, VMs, write-heavy workloads |
The Five Levels in Detail
RAID 0 — Striping, No Redundancy
RAID 0 stripes data across all drives with no duplicate or parity information. You get the full combined capacity and the best sequential throughput, but the failure math is brutal: lose any single drive and the entire volume is gone. Because every added drive increases the odds of that happening, RAID 0 belongs only where data is expendable and easily recreated — video-editing scratch space, temporary render targets, or cache tiers that can be rebuilt from a source of truth.
RAID 1 — Mirroring
RAID 1 writes identical data to two drives. Usable capacity is half of raw, one drive can fail without downtime, and rebuilds are simple block-for-block copies that finish quickly and put little stress on the surviving drive. Reads can be served from either member, so read performance is slightly better than a single drive. This is the default choice for OS and boot volumes — most Dell PowerEdge and HPE ProLiant servers ship with a mirrored pair of drives or M.2 modules for exactly this purpose.
RAID 5 — Single Distributed Parity
RAID 5 stripes data plus one parity block across at least three drives. Usable capacity is N−1 drives, making it the most space-efficient way to get redundancy. The trade-offs: every write requires a read-modify-write cycle to update parity (the classic write penalty), and the array tolerates only a single failure. Rebuilding a failed multi-terabyte drive forces the controller to read every sector of every surviving drive — a window in which a second failure or an unrecoverable read error destroys the array. As a rule of thumb, avoid RAID 5 on arrays built from large nearline drives (roughly 4TB and up); it remains reasonable for smaller arrays of 10K/15K SAS drives or SSDs with a solid backup behind them.
RAID 6 — Dual Distributed Parity
RAID 6 adds a second, independent parity block, so the array survives any two simultaneous drive failures — including a failure during a rebuild, which is precisely the scenario that worries RAID 5 owners. Usable capacity is N−2 drives and the write penalty is higher than RAID 5, so it is rarely chosen for latency-sensitive workloads. Where it shines is bulk capacity: file servers, media archives, surveillance storage and backup targets built from high-capacity 7.2K nearline SAS or SATA drives.
RAID 10 — Mirrored Stripes
RAID 10 stripes data across a set of mirrored pairs, combining RAID 1 protection with RAID 0 speed. Usable capacity is 50%, but there is no parity math: writes are fast, performance barely drops while degraded, and rebuilds are quick mirror copies rather than whole-array parity recomputations. The array survives one failure per mirror pair — potentially several drives, as long as no pair loses both members. For transactional databases, virtualization hosts and anything write-intensive, RAID 10 is the standard answer when the 50% capacity cost is acceptable.
Rebuild Windows: The Hidden Risk
An array is most vulnerable while rebuilding. Parity rebuilds (RAID 5/6) must read every surviving drive end to end while still serving production I/O, and on large nearline drives that can take days. During that window performance drops, the remaining drives run under maximum stress, and any additional failure can be fatal to RAID 5. This is why capacity arrays today default to RAID 6 or RAID 10, and why configuring a hot spare — an installed standby drive the controller can begin rebuilding onto immediately — is cheap insurance for any parity array.
Hardware: Controllers and Drives Matter
Parity RAID performs best on a hardware controller with onboard cache protected by a battery or supercapacitor — think Dell PERC, HPE Smart Array or Broadcom/LSI MegaRAID families. Write-back cache absorbs most of the parity penalty; without cache protection the controller falls back to slower write-through mode. If you run ZFS, Storage Spaces or vSAN instead, use a plain HBA in IT mode rather than a RAID controller. You can find both in our RAID controllers and HBAs collection.
Drive choice matters just as much. Use enterprise-rated drives with time-limited error recovery, and keep interface, capacity and spindle speed consistent within an array — a single slow or desktop-class drive drags the whole volume down and is prone to being dropped by the controller. Our enterprise server hard drives collection carries tested SAS and SATA drives, including Dell- and HPE-labeled options that report cleanly in PERC and Smart Array tooling.
Which RAID Level Should You Choose?
- OS and boot volumes: RAID 1 — simple, fast rebuilds, minimal cost at small capacities.
- Databases, VMs and write-heavy applications: RAID 10 — best write performance and the fastest, safest rebuilds.
- Bulk storage on large drives (file, media, surveillance, backup targets): RAID 6, ideally with a hot spare.
- Smaller arrays of 10K/15K SAS or SSDs where capacity efficiency matters: RAID 5 — acceptable with a verified backup.
- Expendable scratch space only: RAID 0.
Still unsure how many drives to buy or which controller your server generation supports? Our team specs arrays for Dell PowerEdge and HPE ProLiant systems daily — reach out and we will match drives, controller and hot spare to your workload and budget.