Blog
RAID Configurations Explained: RAID 0, 1, 5, 6 and 10 for Servers
RAID decides how a server’s drives work together: how much usable space you get, how fast the array reads and writes, and how many drive failures it can survive. Choosing the wrong RAID level is one of the most common mistakes when setting up a server, and it is hard to change later without rebuilding the array. This guide explains the common RAID configurations, RAID 0, 1, 5, 6 and 10, with the trade-offs in plain language and a quick way to choose.
Key takeaways
- RAID combines drives for performance, redundancy or both. It protects against drive failure, not deletion, ransomware or a failed controller.
- RAID 10 gives the best write performance with redundancy, at the cost of half your raw capacity.
- RAID 6 survives two drive failures, which matters with large drives and long rebuild times.
- A hardware RAID controller with battery- or flash-backed cache improves write performance and protects in-flight data.
- Keep a spare drive and monitor drive health so a failure is replaced before a second one follows.
RAID levels compared
| RAID level | How it works | Minimum drives | Usable capacity | Drive failures survived | Best for |
|---|---|---|---|---|---|
| RAID 0 | Striping, no redundancy | 2 | All drives | None | Scratch space and temporary data only |
| RAID 1 | Mirroring | 2 | One drive | One | Boot drives, small servers |
| RAID 5 | Striping with single parity | 3 | All drives minus one | One | Read-heavy data on smaller drives |
| RAID 6 | Striping with double parity | 4 | All drives minus two | Two | Large file, archive and backup storage |
| RAID 10 | Striped mirrors | 4 | Half the drives | One per mirror pair | Databases, virtual machines, busy applications |
RAID 0: striping for speed
RAID 0 splits data across all drives, so reads and writes are spread out and the array is fast, and you get the full capacity of every drive. There is no redundancy: if any one drive fails, the whole array is lost. Use it only for data that can be recreated, such as render scratch space or temporary caches.
RAID 1: mirroring for simple protection
RAID 1 writes identical copies to two drives. If one fails, the server keeps running on the other, and you replace the failed drive while the mirror rebuilds. You only get the capacity of one drive. It is the standard choice for operating system boot volumes and for small servers with two drive bays.
RAID 5: single parity
RAID 5 stripes data and parity information across three or more drives, so you lose one drive’s worth of capacity and can survive one drive failure. Reads are fast, but every small write needs extra parity work (a write penalty of four disk operations per write), so write-heavy workloads suffer. The bigger concern is rebuild risk: while the array rebuilds after a failure, it has no redundancy left, and rebuilds on large, slow drives can take many hours. For large-capacity drives, RAID 6 is usually the safer choice.
RAID 6: double parity
RAID 6 adds a second parity block, so it survives two simultaneous drive failures and needs at least four drives. You lose two drives’ worth of capacity and writes are slower than RAID 5 (a write penalty of six operations), but you keep protection during a rebuild. It is the sensible default for file servers, archives and backup targets built on high-capacity drives.
RAID 10: mirrored pairs, striped
RAID 10 mirrors drives in pairs, then stripes data across the pairs. It needs at least four drives and gives you half the raw capacity, but it has the best write performance of the redundant levels (a write penalty of two) and rebuilds quickly, because only a mirror copy is needed. It can survive several failures as long as both drives in the same pair do not fail. Choose it for databases, virtualisation hosts and anything sensitive to write latency.
Usable capacity example
With four 1.2TB SAS drives in a server, the usable space works out as follows:
- RAID 0: 4.8TB, with no protection.
- RAID 5: 3.6TB, survives one failure.
- RAID 6: 2.4TB, survives two failures.
- RAID 10: 2.4TB, survives one failure per mirror pair, with the best write performance.
Hardware RAID vs software RAID
Most refurbished business servers include a hardware RAID controller, such as a Dell PERC or an HPE Smart Array card. A controller with battery- or flash-backed write cache can acknowledge writes quickly and protect data in the cache if power is lost, which makes a noticeable difference to RAID 5 and 6 write performance. Software RAID, such as Linux mdadm, Windows Storage Spaces or ZFS, uses the server’s CPU instead and gives you flexibility without depending on a specific controller. Some hypervisors and storage platforms, such as ZFS and VMware vSAN, expect drives to be passed through without hardware RAID, so check what your software needs before you configure the controller.
How to choose a RAID level
- Two drive bays or a boot volume? RAID 1.
- Databases, virtual machines, email or busy applications? RAID 10 on SAS drives or SSDs.
- Large file shares, archives or backup storage? RAID 6 on high-capacity drives.
- Mostly read workloads on a small number of smaller drives? RAID 5 is acceptable, with good backups.
- Data you can recreate? RAID 0, never for anything important.
Whatever you choose, use drives of the same size and type in an array, keep a hot or cold spare, and monitor drive health through the controller or the server’s management interface, such as Dell iDRAC or HPE iLO.
RAID is not a backup
RAID keeps a server running when a drive fails. It does not protect you from accidental deletion, ransomware, corruption, a failed controller or a lost server. Keep separate backups, ideally including one copy off-site, and test that you can restore them.
Frequently asked questions
What is the best RAID configuration for a server?
For most business servers, RAID 10 for performance-sensitive workloads and RAID 6 for large-capacity storage. RAID 1 is the right choice for boot drives and two-drive servers.
Is RAID 5 still safe to use?
RAID 5 works, but it only survives one drive failure and has no protection during a rebuild. With large drives and long rebuilds, RAID 6 or RAID 10 is safer.
Can I mix SSDs and hard drives in one RAID array?
You should not. Keep each array to drives of the same type, size and speed. You can run separate arrays, for example SSDs in RAID 1 for the operating system and hard drives in RAID 6 for data.
How many drive failures can RAID 10 survive?
At least one, and more as long as no two failed drives are in the same mirror pair.
Need drives or a server for a new array? Browse our refurbished server hard drives and SSDs, our storage servers and storage systems, or tell us about your workload and we will suggest a RAID layout and drive configuration.