How Much Space Do You Actually Get From RAID?

Say it before anything else: RAID is not a backup

It protects against a drive failing. That's the entire scope.

It does nothing about deleting the wrong folder, ransomware, a corrupt file written correctly to every disk, a dead controller, theft, fire or flood. In every one of those cases the redundancy works exactly as designed and replicates the problem faithfully across all your drives.

Redundancy is uptime insurance, not data insurance. An array with no separate backup is one mistake from total loss no matter which level you picked.

With that established, the arithmetic.

Six 4 TB drives, five different answers

Level	Usable	Efficiency	Drives that can fail
RAID 0	24 TB	100%	0
RAID 5	20 TB	83.3%	1
RAID 6	16 TB	66.7%	2
RAID 10	12 TB	50%	1
RAID 1	4 TB	16.7%	5

Same 24 TB of hardware. A sixfold spread in what you get to use.

Price it, don't eyeball it

"RAID 6 costs you another whole drive" sounds like a lot. Put a price on it instead — at $100 a drive:

Level	Usable	Cost per usable TB
RAID 5	20 TB	$30.00
RAID 6	16 TB	$37.50

Twenty-five percent more, for double the fault tolerance. That reads very differently from "you lose another drive", and it's the comparison that should drive the decision.

Why the premium is usually worth paying: when a drive dies in RAID 5, every remaining drive must be read completely to reconstruct it. A second failure during that window loses everything. With large modern drives that window is long — which is a separate calculation, and the reason to look at rebuild time before settling on a layout.

Why RAID 1 is so expensive

Every drive holds a complete copy, so usable capacity is one drive's worth however many you add.

Six 4 TB drives in RAID 1 is 4 TB usable at $150 per usable terabyte — five times RAID 5 on identical hardware. What you buy is extreme fault tolerance: five of six drives can die and the data survives.

That trade makes sense for two drives holding something critical. Beyond that it almost never does, which is what RAID 10 is for — mirrored pairs striped together, keeping half your capacity instead of a sixth.

Two things that catch people

The OS will report less, and nothing is missing. Drives are sold in decimal terabytes (a trillion bytes). Windows reports tebibytes (1,099,511,627,776 bytes) and calls them TB anyway. It's a fixed 9.1% gap — 20 usable TB shows as 18.2. Formatting and filesystem metadata come off after that.

RAID 10 with an odd drive count wastes one. Five 4 TB drives gives 8 TB — exactly what four would give. The fifth contributes nothing to capacity and is only useful as a hot spare. The tool floors it rather than reporting half a drive of storage.

The fault-tolerance number is the guaranteed one

RAID 10 might survive several failures, if they happen to land in different mirror pairs. It might also die on the second drive if both hit the same pair.

So the tool reports 1, because that's what you can plan against. Hoping the failures are politely distributed isn't a storage strategy.

RAID 1 is the genuine exception — it really does survive until one drive is left.

Fewer big drives, or more small ones?

On cost per usable terabyte, usually fewer and bigger. Large drives are generally cheaper per terabyte, and a parity level gives up a fixed number of drives rather than a fixed share — so parity costs proportionally less in a bigger array.

But capacity isn't the only axis. Bigger drives take longer to rebuild, and a longer rebuild is a longer window in which a second failure is fatal. The level that looked adequate on a small array may not be on a large one.

That's the trade worth checking before you buy, not after.