Disks are among the components that will fail sooner or later. If a server has only one disk, its failure means a complete outage and possibly lost data. RAID arranges several disks so the system keeps running when one or more of them fail. But RAID levels differ greatly in data protection, speed and usable space.
You have probably heard of RAID 0 as well. Data is spread across the disks and speed goes up, but there is no redundancy at all, and if any disk fails, all the data is lost. That makes it unsuitable for servers.
RAID 1: Mirroring
- Structure: two disks that both hold exactly the same data.
- Fault tolerance: one disk.
- Usable space: half the total; with two 1 TB disks, you get 1 TB.
- Use case: server operating system disks and small servers.
RAID 5 and RAID 6: Parity
At these two levels, data is spread across all disks along with parity information, so a failed disk’s data can be rebuilt from the others.
- RAID 5: at least three disks, tolerates the failure of one disk; the space of one disk is used for parity.
- RAID 6: at least four disks, tolerates the failure of two disks; the space of two disks is used for parity.
For example, with six 4 TB disks, RAID 5 gives about 20 TB of usable space and RAID 6 about 16 TB. The weakness of both levels is slower write speed because of parity calculations.
With high-capacity disks, rebuilding an array after a failure can take a long time, and all the disks are under heavy load during the rebuild. If another disk fails or returns a read error during that window in RAID 5, the array is lost. That is why RAID 6 is the safer choice for high-capacity mechanical disks.
RAID 10: Protection and Speed
A combination of RAID 1 and RAID 0: disks are mirrored in pairs, and data is striped across those pairs.
- Minimum disks: four, always an even number.
- Fault tolerance: at least one disk, and more if the failures are in different pairs; but losing both disks in one pair means losing the array.
- Usable space: half the total.
- Use case: databases and virtualization hosts that need high read and write speed.
Which One Should You Choose?
- Server operating system: RAID 1
- File and archive servers with high-capacity disks: RAID 6
- Databases, virtual machines and heavy write workloads: RAID 10
- Backup storage: usually RAID 6, for its balance of space and protection
A few practical points: use a hardware RAID controller with protected cache, or a proven software solution such as ZFS in Proxmox. If possible, keep a Hot Spare disk so rebuilding starts immediately, and use disks of the same size and type within an array. Monitor the array as well. An array sitting in Degraded state without anyone knowing is one step away from disaster.
RAID Is Not a Backup
RAID protects only against disk failure. If a file is deleted by mistake or ransomware encrypts your data, that change is applied instantly to every disk. A controller failure, fire or server theft also wipes out the whole array at once. You need RAID for keeping work going and backup for getting your data back.
Checklist
- Each server’s RAID level matches its workload.
- RAID 5 is not relied on for high-capacity mechanical disks.
- The RAID controller has protected cache and up-to-date firmware.
- A Hot Spare or spare disk is ready.
- Array status alerts are enabled.
- An independent, tested backup exists alongside RAID.
ITPlus724 engineers can review the RAID setup on your servers and design and implement the right level for new servers and storage.