I earn commissions when you sign up or buy through the affiliate links on this page, at no extra cost to you.
Homelab · Guide · By Mohammed Almuhanna · Updated
RAID Levels Explained: 0, 1, 5, 6, and 10
RAID combines several drives into one array for redundancy, performance, or capacity. One thing first, because people get it wrong constantly. RAID is not a backup. It covers a drive dying, nothing else. It will not save you from deleting a file, from ransomware, or from a fire that takes the whole machine. You still need real backups. I run SHR-1 on five IronWolf Pro 12TB in a Synology at home, which is the RAID 5 class, single parity. I stand by it for a home NAS, but only because I keep a real offsite copy. Without that backup I would not run single parity, and neither should you. Here is how each common level actually works and which one I would pick.
RAID 0 (striping)
RAID 0 splits your data across every drive in the array. There is no redundancy at all, so each drive holds a unique slice of the data.
- How many can fail: none. One drive dies and the whole array is gone, every byte on every disk.
- Usable capacity: all of it. Two 4 TB drives give 8 TB usable.
- Performance: fast. Reads and writes spread across all drives, so throughput scales with the drive count.
- My take: RAID 0 is a toy for scratch data, nothing more. A video editing cache you can rebuild, that kind of thing. The more drives you add, the more likely the array is to die, because any single failure kills it. Never put anything you care about on RAID 0.
RAID 1 (mirroring)
RAID 1 writes identical copies of your data to two or more drives. Every drive in the mirror holds the same content.
- How many can fail: one, in a standard two-drive mirror. The array keeps running on the surviving copy.
- Usable capacity: half. Two 4 TB drives give 4 TB usable.
- Performance: reads are quick since either drive can serve them. Writes go to both copies.
- My take: if you have exactly two drives, run RAID 1 and stop thinking about it. It is the simplest thing that works, no parity math, and a rebuild just copies the surviving disk onto the new one. You pay for two drives and get the capacity of one, and that is the whole trade.
RAID 5 (single parity)
RAID 5 stripes your data across three or more drives and adds one block of parity per stripe, spread across the drives. The parity lets the array rebuild the contents of any single failed drive. This is the single-parity class, the same idea as Synology's SHR-1, which is what I run at home.
- How many can fail: one. A second failure before the rebuild finishes loses everything.
- Usable capacity: (n − 1) drives. Four 4 TB drives give 12 TB usable.
- Performance: good read speed. Writes carry a parity calculation overhead.
- My take: this is the level I actually use, and I will defend it for a backed-up home array. It gives the best capacity of any parity level. But I only run it because I keep a real offsite backup, and that is what you are paying for. The risk lives in the rebuild. When a drive fails, rebuilding onto its replacement reads every other drive in full, which takes many hours on big modern drives, and during that window the array has zero redundancy. A second drive failure or one read error in that window takes the whole array down. So on a large array of large drives, single parity is risky, and if the data is not backed up I would not touch it. See why single parity is risky on big drives.
RAID 6 (double parity)
RAID 6 works like RAID 5 but keeps two independent parity blocks per stripe across four or more drives. That second parity block is what makes it the safer choice for large arrays. ZFS calls the same idea RAIDZ2, and Synology calls it SHR-2.
- How many can fail: two. The array still rebuilds even if a second drive dies during the first rebuild.
- Usable capacity: (n − 2) drives. Six 4 TB drives give 16 TB usable.
- Performance: reads are similar to RAID 5. Writes cost a bit more because two parity blocks are computed.
- My take: double parity is the sane default for any big array of big drives. The second parity disk covers the exact failure that sinks RAID 5, a second fault during that long rebuild window. You give up one more drive of capacity for it. On a large array, that trade is worth it every time.
RAID 10 (1+0, mirrored stripes)
RAID 10 combines mirroring and striping. Drives are paired into mirrors, then data is striped across those mirrored pairs. It needs at least four drives, in even numbers.
- How many can fail: at least one, and often more. The array survives a drive in each mirror failing, but loses everything if both drives in the same mirror die.
- Usable capacity: half. Four 4 TB drives give 8 TB usable.
- Performance: fast, and it rebuilds quickly. A rebuild just copies from the surviving mirror, with no parity to recompute across the whole array.
- My take: RAID 10 trades capacity for speed and a fast, low-risk rebuild. You only get half your raw space, same as a mirror, so it is an expensive way to store bulk data. But for databases and VM storage, where write speed matters and you want the rebuild over with quickly, it is the right call. For a media NAS it is a waste of disks.
The levels side by side
| Level | Min drives | Drives that can fail | Usable capacity | Relative speed | Typical use |
|---|---|---|---|---|---|
| RAID 0 | 2 | 0 | All (n) | Fastest | Scratch and cache, no important data |
| RAID 1 | 2 | 1 | Half (n / 2) | Moderate | Small, simple, safe setups |
| RAID 5 | 3 | 1 | n − 1 | Good | Capacity on smaller or less critical drives |
| RAID 6 | 4 | 2 | n − 2 | Good | Large arrays of large drives |
| RAID 10 | 4 | 1 or more | Half (n / 2) | Fast | Databases and VM storage |
Which should you pick?
Here is how I would call it. Two drives, run RAID 1. A big array of large drives where capacity matters, run RAID 6 so it survives a failure during the rebuild. Speed-sensitive work like databases and VMs, run RAID 10. Single parity (RAID 5 or SHR-1) is fine for a home array, but only if you keep a real offsite backup, which is exactly why I run it myself. Without that backup, do not put single parity on large modern drives, because the rebuild risk turns one dead disk into a likely total loss. To see exact usable capacity and fault tolerance for your own drive count and size, use the RAID storage calculator.
Common questions
Is RAID a backup?
No, and this is the one people get wrong. RAID covers a drive dying, nothing else. It will not save you from deleting a file, from ransomware, or from a fire that destroys the whole machine. You still need real backups, ideally an offsite copy. RAID buys you uptime and a safety margin, not a backup.
Which RAID level is best for a home NAS?
For two drives, RAID 1 is the simple safe choice. For a larger array of large drives where capacity matters, RAID 6 is safer because it survives a second failure during the rebuild. Single parity (RAID 5 or SHR-1) is fine if you keep a real offsite backup, which is what I run at home. RAID 10 is the pick for speed-sensitive work like databases and VM storage.
Why is RAID 5 risky on large modern drives?
When a drive fails, rebuilding onto the replacement reads every other drive in full, which takes many hours on big drives. During that window the array has no redundancy, so a second drive failure or a single read error loses everything. RAID 6 covers exactly that risk with a second parity block. Single parity is still fine for a home array if you keep a real offsite backup, but on large unbacked drives I would not run it.
How much usable capacity does each level give?
RAID 0 gives all of it, RAID 1 gives half, RAID 5 gives (n − 1) drives, RAID 6 gives (n − 2) drives, and RAID 10 gives half. So four 4 TB drives give 8 TB in RAID 0, 12 TB in RAID 5, and 8 TB in RAID 10.