RAID Levels Explained for Business Servers: RAID 1, 5, 6 and 10
When you configure a server, the storage layout is one of the decisions that is hardest to change later. RAID (redundant array of independent disks) combines several drives so that the server can survive a drive failure, run faster, or both. The right level depends on what you store, how much you can afford to lose, and how fast the workload needs to be. This guide compares the levels most often used in business servers.
First: RAID is not a backup
RAID keeps a server running when a drive fails. It does not protect against deleted files, ransomware, controller failure, fire or human error, because those affect every drive in the array at once. Keep separate backups, preferably stored off the server and off site, whichever RAID level you choose.
RAID 0 — speed, no protection
Data is split across drives. It is fast and uses all capacity, but if any one drive fails, all data is lost. It is suitable only for temporary or easily rebuilt data, not for business records.
RAID 1 — mirroring
Two drives hold identical copies. If one fails the other continues. Usable capacity is that of one drive (50% of the total), and it is simple and reliable. It is a common choice for the operating system drives of a server.
RAID 5 — striping with single parity
Needs at least three drives. Capacity of one drive is used for parity, so usable capacity is the total minus one drive, and the array survives one drive failure. It is capacity-efficient, but writes are slower because parity must be calculated, and during a rebuild after a failure the remaining drives are under heavy load. With very large hard drives, rebuilds can take long enough that a second failure during the rebuild is a real risk.
RAID 6 — striping with double parity
Needs at least four drives. Two drives' worth of capacity is used for parity, so it survives two simultaneous failures. It costs more capacity than RAID 5 and writes are slower still, but it is a safer choice for arrays of large drives that hold important data.
RAID 10 — mirrored pairs, then striped
Needs at least four drives. Drives are mirrored in pairs and the pairs are striped. Usable capacity is 50% of the total, it gives strong read and write performance, and rebuilds are fast because only one mirror needs to be copied. It is often chosen for databases and virtual machine storage where performance and quick recovery matter more than capacity.
Quick comparison
- RAID 1: 2 drives minimum, usable capacity of one drive, survives one failure.
- RAID 5: 3 drives minimum, usable capacity of all drives minus one, survives one failure.
- RAID 6: 4 drives minimum, usable capacity of all drives minus two, survives two failures.
- RAID 10: 4 drives minimum, usable capacity of half the drives, survives one failure per mirrored pair.
Other points to specify when buying
- Hardware or software RAID. A dedicated RAID controller with cache protection (a battery or capacitor) is common in servers; ask which controller model is included and whether the cache is protected against power loss.
- Drive type. SSDs and hard drives behave differently in rebuild time, performance and endurance; state the workload so the right drive class is chosen.
- Hot spares. An extra drive that takes over automatically shortens the time the array runs without protection.
- Monitoring. Confirm how drive failures will be reported, so a failed drive is replaced before a second one fails.
For the other parts of the build, see our guides to sourcing enterprise servers, ECC server memory and server hardware standards.
Summary
For most business servers, RAID 1 for the operating system and RAID 10 or RAID 6 for data covers the common cases, with RAID 5 reserved for smaller arrays where capacity efficiency matters. The decision should be written into the specification alongside the drive type, controller and backup plan. If you are sourcing servers, you can browse storage products or request a quote with your capacity and performance requirements.
Related products: 2U Rack Server, Tower Server
